Home

Office energy saving potential through component based automation

1. 568 562 i 8 8 a x TB rium J CTT TTT HEN E i Figure 4 1 M ap of the offices and kitchen area on the east side of the fifth floor The east side of the fifth floor consists out of four offices and a kitchen area The offices give room to one or more employees Figure 4 1 showsthe layout of the workplaces and light fixtures in those offices as well asthe kitchen area 32 73 Figure 4 2 shows the layout of the north wing which contains the other half of the environment controlled by the system The right side of the hallway in Figure 4 1 connects to the bottom of the hallway in Figure 4 2 The layout of the workplaces and light fixtures in the north wing is depicted in Figure 4 2 The north wing houses four offices and one meeting area Unequipped Wim i Unequipped Cm Figure 4 2 Map of the offices and meeting area on the north wing of the fifth floor 4 1 1 Kitchen area The kitchen area is a common social area where office workers mostly meet during lunch breaks to use the microwave or to get their food from the fridge The kitchen has the following types of appliances that can be controlled 1xfridge 1xwater boiler 1xmicrowave 1xcoffee machine 33 73 4 1 2 Meeting area The meeting area is a designated office for holding discussions and presentations The meeting
2. 66 73 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 SPARSITY TECHNOLOGIES Sparsity technologies DEX high performance graph database visited 02 05 2013 http www sparsity technologies com dex_tutorials4 name Architecture SPARSITY TECHNOLOGIES Sparsity technologies DEX high performance graph database visited 02 05 2013 http www sparsity technologies com events BLOGSPOT NuvolaBase s Blog XDGBench 3rd party benchmark results against graph databases visited 18 07 2013 http nuvolabase blogspot sg 2013 04 xdgbench 3rd party benchmark results html JAKUT Graph Databases Vytautas Jakutis visited 18 07 2013 https jakut is 2013 05 12 graph databases BULBFLOW The Bulbs Community a Python Framework for Graph Databases Bulbflow visited 18 07 2013 http bulbflow com community GOOGLE Google Groups visited 18 07 2013 https groups google com forum forum aureliusgraphs GOOGLE 99 HyperGraphDB Google Groups visited 18 07 2013 https groups google com forum forum hypergraphdb GOOGLE Status of Replication Google Groups visited 18 07 2013 https groups google com forum topic hypergraphdb fBBVD_Rh_eA HYPERGRAPHDB HypergraphDB A Graph Database visited 18 07 2013 http www hypergraphdb org blog entry http www blogger com feeds 1
3. 7 university of faculty of sptar aa roninsen and natural sciences C 7 8 g 3 N S Office energy saving potential through component based automation a design and implementation M arcel Jillings Sijmon Heitmeijer University of Groningen University of Groningen Software Engineering and Distributed Systems Software Engineering and Distributed Systems supervised by Alexander Lazovik University of Groningen The Netherlands Alexandru Telea University of Groningen The Netherlands August 27 2013 Version 1 0 Abstract One of the focus points of the University of Groningen in teaching and research is energy and sustainability In 2012 two PhD students from the Distributed Systems research group proposed a prizewinning plan to make the Bernoulliborg building sustainable in terms of waste disposal and the use of water and energy The proposed plan is being executed in the course of the year 2013 This thesis is part of the execution of the plan s energy savings solution that aims to save at least 7 796 107 514 kWh annually through automatically controlling HVAC systems lights computers appliances and providing energy consumption tracking A new automated system had been designed and installed in the north wing on the fifth floor of the Bernoulliborg building that is capable of monitoring and influencing the energy consumption of the offices and common areas Plugwise devices have been installed after considering multiple
4. Reliability varies e g rotating the office chair Does not detect computer usage e g user is making notes in front of the computer Price 65 1g4B custom built excluding development and build costs 29 30 41 42 Other sensors Several other sensors were considered but dropped for consideration for the actual approach Radio frequency identification RFID near field communication NFC or WiFi tags due to the accompanying requirement to use badges Badges can be lost forgotten and are not ideal in an open building such as an university because of the amount of visitors and students 19 73 Door sensors due to inability to detect multiple persons in an office and due to the inaccuracy e g standing in the door opening to simply wave to a colleague Cameras due to privacy concerns Mobile phone tracking software due to privacy concerns and similar disadvantages as the previously mentioned tags 3 2 2 Computer The computer is the main center of activity in an office Per computer in an office there are two variables to be measured to determine context and to influence the energy consumption Energy consumption where either the computer and monitor are measured for determining context separately or combined A baseline measurement of the energy consumption compared to later measurements can measure how effective the solution is Computer usage which consists of the idle time i e the computer is po
5. SMS The sleep management server component and the SAGW computer client were developed together The computer client runs on any workstation and exposes certain abilities from the computer to the Green M ind system Besides a value to indicate whether the computer is active or idle is the possibility to put the computer to sleep hibernate on or off included in these abilities The workstations and their states are managed by the SMS When a state changes of a computer client then the SM S will communicate with the context component by sending out a notification that the context variables need to be updated accordingly Other components e g the orchestrator can instruct the SM S to change the state of a given computer but also request information about the computer from the SM S if necessary The SAGW computer client was developed in C as opposed to Java for the other components because when C is used for development there is no need to install the Java dependency on each workstation This provides a faster easier and better manageable solution that also adheres to the ICT security policy of the university of Groningen Dashboard To raise awareness and decrease energy consumption as much as possible it is important to provide feedback or to create a social element All the visualizations of the Green Mind system are handled by the dashboard Employees can view their consumption online or through a mobile app The dashboard and mobile
6. status error data code 123 message PlugwiseDevice is not set to an instance of an object Figure 4 8 JSON object of a typical error response In the case of Figure 4 7 the component correctly processed the request and returned all the available data Figure 4 8 shows a JSON object for when the request fails or an error occurred Instead of returning any of the requested data the JSON object contains a specific error code and a message that describes the problem The code field is internally unique for a component e g code 1 might mean invalid parameter for one component and could not find device for another but cannot refer to more than one specific error The status field can have one of the three states listed in Table 4 1 Success All went well and optional data was returned fail There was a problem with the submitted data parameters error An error occurred while processing the result Table4 1 The three available values for the status field in a response J SON object 4 5 Discussion The Green Mind system has been implemented in the course of about six months and the required components are accounted for However at the moment of writing the system as a whole is only partially operational which can only be solved by further development of the partial finished components The components in question are the SAGW of which basic functionality is operational but only as a temporary standalone com
7. St h Gig yOR Sy amp eNC a 2yyA LEN AGES Sy amp aNjC a 2yy u ZNOCEABR 25 06 2013 http www monnit com products wireless sensors seat occupancy php ARPCI Wireless Industrial Seat Occupancy Sensor AA Powered visited 25 06 2013 http www arpci com industrial sensors 723 ind091a 0so html PLUGWISE Circle Plugwise visited 25 06 2013 http www plugwise com nl idplugtype f node 113 ENOCEAN ALLIANCE EnOcean Alliance Products Enabled by EnOcean visited 25 06 2013 http www enocean alliance orq en products eltako fsr61 va PREISSUCHM ASCHINE ELTAKO ELTAKO FBH63AP rw Funk Bewegungs Helligkeitssensorreinweiss 30000852 Preisvergleich G nstig kaufen bei Preissuchmaschine de visited 25 06 2013 http www preissuchmaschine de in Diverses Diverse E ELTAKO ELTAKO FBH63AP rw Funk Bewegungs Helligkeitssensorreinweiss 30000852 html ENOLUZ M otion Sensors Thermokon SR M DS BAT Wireless Ceiling M ultiSensor 3609 visited 25 06 2013 http www enoluz com thermokon wireless ceiling multisensor 3600rdm p 455 html DIGIKEY EKM B1101112 Panasonic Electric Works 255 3066 ND DigiKey visited 25 06 2013 http www digikey com product detail en EKMB1101112 255 3066 ND 2601860 64 73 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 M Y KNX SHOP Eltako Funk Helligkeitssensor
8. Fixtures already contain a PIR sensor and cannot be manually turned on or off Figure 5 1 Energy consumption summary of office lights A summary of the energy consumption of the lights in the offices is given in Figure 5 1 Important items to note are that the monitored offices have at least two and at most three separate light fixtures These fixtures are already controlled by a PIR sensor in the office This PIR sensor could not be used by the Green M ind system and thus the Plugwise device was installed on top of the fixtures and the PIR sensor This means that even though the Plugwise device is turned on the PIR sensor still needs to register movement for the lights to work It is not possible for the office workers to turn lights on or off manually The effects of the PIR sensor are also represented in the two graphs of Figure 5 1 Most noticeable at night and in the weekends where the lights do not consume any energy but also on regular weekdays where the lights stay on all day even when the sun is active because there is movement in the office A single fixture consumes around 75 watt when turned on This means that the lighting of a single office consumes 150 225 watt per hour for around eight hours This equals to around 1200 1800 watt per day that is simply being wasted on a sunny day The PIR sensor alone is not enough to save energy With an average of 4 5 hours 203 5696 of the eight hours of work time per office of sunlight per day
9. www rabbitma com getstarted html APACHE Powered By Apache Kafka Apache Software Foundation visited 29 07 2013 https cwiki apache org confluence display KAFKA Powered By NABBLE RabbitMQ Mailing List Archive visited 29 07 2013 http rabbitmq 1065348 n5 nabble com STACKOVERFLOW Questions containing rabbitmq Stack Overflow visited 29 07 2013 http stackoverflow com search q rabbitmg STACKOVERFLOW Questions containing hornetq Stack Overflow visited 29 07 2013 http stackoverflow com search q hornetg JBOSS HornetQ JBoss Community visited 29 07 2013 https community jboss org en hornetq view discussions STACKOVERFLOW Questions containing kafka Stack Overflow visited 29 07 2013 http stackoverflow com search q kafka APACHE kafka users mailing list archives visited 29 07 2013 http mail archives apache org mod_mbox kafka users 201307 mbox browser RABBITM Q RabbitM Q Documentation visited 29 07 2013 http www rabbitmg com documentation html RABBITM Q RabbitM Q Java Client API Guide visited 29 07 2013 http www rabbitmg com api JAY KREPS NEHA NARKHEDE AND JUN RAO Kafka A distributed messaging system for log processing In Proceedings of 6th International Workshop on Networking M eets Databases NetDB Athens Greece 2011 Dht hd pj pjj a Sid 3S3 dio 26 LAA Ml WO aw ooAa v Ba v ttime Traffic M aps Pivotal P O V visited 01 08 2013 http blog gopivotal com case stu
10. Another way isto allow a select group of users to temporarily override the activity detection mechanism The computer software sensor can also be equipped with the ability to turn a computer off to sleep and to hibernate In case of laptops the portability aspect does not necessarily cause concerns for the computer software sensor to function As long as the laptop remains reachable by the other components of the system it can still be used to turn the laptop to sleep hibernate or off The energy consumption measurements for a laptop remain a point of interest with or without the computer software sensor e g laptop is charged elsewhere or is not plugged in at the office because of remaining battery life Advantages Detect computer usage Easy installation at no additional hardware costs Unobtrusive placement Updatable to incorporate new features Disadvantages Requires constant activity Easyfor users to avoid e g by killing the process of the software sensor Price j B excluding development costs Ultrasonic sensor An ultrasonic sensor generates sound waves and evaluates their echos to measure the distance to an object The sensor can be placed in front of the user faced to the office chair Desk usage can be detected by measuring the difference between an empty and occupied office chair Advantages Detect desk usage Does not require movement Disadvantages Obtrusive placement in front of user
11. Haris Doukas Konstantinos D Patlitzianas Konstantinos latropoulos John Psarras Intelligent building energy management system using rule sets Building and Environment Volume 42 Issue 10 October 2007 Pages 3562 3569 J KW Wong H Li S W Wang Intelligent building research a review Automation in Construction Volume 14 Issue 1 January 2005 Pages 143 159 63 73 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 Omar M Al Rabghi Mohammed M Akyurt A survey of energy efficient strategies for effective air conditioning Energy Conversion and Management Volume 45 Issues 11742 July 2004 Pages 1643 1654 H W Kua S E Lee Demonstration intelligent buildinglYa methodology for the promotion of total sustainability in the built environment Building and Environment Volume 37 Issue 3 M arch 2002 Pages 231 240 J Han Y Jeong and I Lee Efficient building energy management system based on Ontology inference rules and simulation in International Conference on Intelligent Building and M anagement vol 5 2011 Pages 295 299 Georgievski IIche and Degeler Viktoriya and Pagani Giuliano Andrea and Nguyen Tuan Anh and Lazovik Alexander and Aiello M arco Optimizing Energy Costs for Offices Connected to the Smart Grid In IEEE Trans Smart Grid 3 4 2273 2285 Year 2012 GREENERBUILDINGS Approach GreenerBuildin
12. about the potential cumulative energy savings made through the system rather than the potential energy savings per device e g computers lighting Despite the scientific interest in improving the energy consumption through BM S systems Nguyen et al 24 describe in a survey about energy intelligent systems that the three main energy consuming subsystems in buildings i e HVAC lighting and office equipment also draw the attention of numerous studies From the researched studies some focus on one subsystem only while others try to save energy for two or even three subsystems The majority 22 out of 32 of the reviewed studies focus only on one subsystem whereas this thesis focuses on two subsystems Six studies focus on HVAC and lighting three focus on HVAC lighting and power plugs and two others focus on lighting and power plugs only iSpace 25 and Intelligent Buildings by Davidson et al 26 are the two projects that only focus on lighting and power plugs similar to the focus of this thesis The work presented in 25 differs with this thesis because although it can be used for energy savings it is focused on autonomous environments and on student bedrooms rather than offices In addition the energy consumption is not monitored Although Davidson et al do the same conceptually 26 the system in this thesis uses a different implementation For example 26 uses a Multi Agent System i e each agent is linked to and manages a p
13. month year hour minute isweekend and isweekday in their evaluations Through the context variables the rules can act immediately upon changes from the sensors and actuators In Figure 4 4 an example is depicted of a simplified rule whereas Figure 4 5 depicts an example of an extended rule To distinguish between a simplified and an extended rule the rule has to start either a or an misc date get hour gt 20 amp amp computers 127 0 0 1 get is idle true gt computers 127 0 0 1 set state off Figure 4 4 Example of a simplified rule to turn a computer off when it is idle and it is after 20 00 O Weather weather new Weather if weather getW eatherFor Groningen Weather Cloudy 1 lights LIGHT_1 set state on Y Figure 4 5 Example of an extended rule to turn a light on when the weather is cloudy The simplified rules are based on a predicate the expressions before the first gt that should evaluate to true before the remaining part the expressions after the first gt of the rule is evaluated and executed The variables that are set in the remaining part of the rule are immediately sent to the orchestrator after the rule is evaluated and executed Figure 4 4 and 4 7 only show examples of a single device but it is also possible to create a rule for a whole group e g a rule for all computers Because the simplified rules are using a predicate to determine whether or not the rema
14. the inserts are coming in at a higher rate than the database can process it will eventually come to a stop The read performance is also important as the components of the system rely on the sensor measurements The performance of reads and writes simultaneously is important too to ensure that all system components continue to function Documentation The documentation of the database can address topics like tutorials how to setup the database drivers language bindings and tutorials A database is not considered without well written and complete documentation as there are too many uncertainties and risks involved Community An important aspect of a database is the community evolving around it Without a community it is often hard for users to ask questions and find tutorials about the product A community is also important to ensure that development continues with input from the users e g suggest new features or submit bug reports Price A free or open source database is preferred because of the open character of the system This includes not limiting the amount of possible nodes for the database Usage The usage or adoption rate is an important aspect of a graph database It affects aspects such as development e g continue with input from the users or the possibility to add patches or functionality in open source software stability e g it is unlikely a company is going to use unstable software in production and support e g more
15. too expensive when using Plugwise devices Each device is unique but continuous monitoring is not always necessary It is best to group devices together where possible An automated building management system such as the proposed Green M ind system should be able to handle the repositioning of sensors and actuators and be installed in phases to optimize the cost benefit ratio The first goal of lowering the energy consumption of lighting by 2596 can be reached by measuring the available daylight and controlling the lights accordingly The current use of motion sensors in offices to control the lighting is insufficient and leaves much room for improvement The estimated energy savings range from 36 for corridor lighting to 56 for office lighting The second goal of lowering the energy consumption of electronic devices by at least 2596 can only be partially satisfied due to the presence of research workstations that cannot save energy through simple automation In the category of electronic devices the potential energy savings with the Green M ind system were 56 for the coffee machine 2896 for the water boiler 2296 for a regular workstation 996 for the fridge and 796 for the microwave M ost of the potential energy savings are below the said goal of 2596 but combined they average a potential energy saving of 27 596 However when the excesses of the research workstations that are operated 24 hours a day and 7 days a week are taken into account t
16. 30 38 or prefabricated 36 37 Light sensor A light sensor can detect light intensity in low light conditions indoor high light conditions outdoor or both Asthe light intensity measured highly depends on the location of the light sensor e g near a window placing multiple light sensors in a shared office with different desk locations can improve the accuracy of influencing the energy consumption based on the light intensity Advantages Determine whether lights should be switched on or off Available as multi sensor light intensity and PIR Disadvantages Minimum acceptable light intensity level depends on the user Interference with lights being switched on Highly dependent on location Price 55 HMj B custom built excluding development and build costs 29 30 40 or prefabricated 36 37 39 18 73 Computer software sensor A computer software sensor can detect input such as mouse and keyboard activity of the user As soon as the mouse is moved the scroll wheel used or a key is pressed activity is registered by the computer software sensor The amount of time after the last registered activity can be used to determine the idle time of the computer However using only the input devices to determine if a computer is idle may not be enough For example consider a complex program that requires several hours to complete One way to solve this problem is by also monitoring the CPU and GPU usage of a computer
17. 7 years at most Savings should be high enough to make the purchase and installation of such a device appealing 1 2 3 Planned solution In order to reach the envisioned goals an automated system needs to be able to influence the energy consumption of lights and electronic devices in the existing offices The system also needs to collect the consumption information of the individual devices for research purposes and to create user awareness to encourage even more energy savings by the users themselves The plan is to reach the goals by installing remotely accessible sensors and actuators in the offices to provide an automated system with all the necessary data to base decisions on and to make it possible to set non essential devices in standby mode or to turn them off completely The governing system itself consists of multiple subcomponents Each component performs different tasks and has their own dedicated goal e g collecting the consumption measurements Components work together to reach their common goal reducing the energy consumption in the offices and to share useful information with the other components so that both the raw sensor data and other processed information is accessible to each component To make the management of these components easier the components communicate either through a REST interface or a message queue which allows for loose coupling between the components REST is the dominant web application programming interfac
18. Bus Fyydizf pt y Briyre J S BH 2T Bay eS Nj CONKKS t OND2XSNKKS Fyydbt Y 2y d N d ys HN estimated to be Bro RRA dej NjdyR Bin 6 Y KS Baby eS NAHE The solution of combining the coffee machine and microwave would mean that the microwave adheres to the powering schedule of the coffee machine because the coffee machine needs to be active the longest per day For the water boiler the envisioned schedule is the same as the coffee machines so the boiler can also be added to the same Plugwise device to save costs Combined the appliances in the kitchen consume 1359 13 kWh annually and have an estimated energy saving potential of 474 87 kWh This means an estimated potential energy saving of 3596 combined for all the devices in the kitchen 34 when the savings of the fridge are left out of the equation or 32 when both the fridge and the microwave are left out Either way the microwave is best included with the coffee machine and water boiler to at least also save some energy 57 73 Chapter 6 Conclusion The Green Mind system that was designed and partially implemented for this thesis showed that previous research conducted by the Distributed Systems group at the University of Groningen also scales to a larger uncontrolled office environment The scale was not expected to be an issue but so was the cost of the system As it turned out there is a lot of potential energy to be saved but tracking consumption of every single device is
19. Introduction This chapter introduces the background and research goals for this thesis about potential energy saving in offices through the use of a component based building automation system First in Section 1 1 the context of the research domain is provided in order to create a better understanding of the origin and the reasons for this research After clarifying the domain the hypothesis and research goals are given in Section 1 2 along with a summary of the used approach and expectations Finally in Section 1 3 the relevance and contributions of this thesis are listed before giving an overview of the further structure of this thesis in Section 1 4 1 1 Context This thesis is conducted as part of multiple research projects related to energy and sustainability at the Distributed Systems research group at the University of Groningen Energy and sustainability is one of the focus points in teaching and research of the university 3 In 2012 Tuan Anh Nguyen and Faris Nizamic two PhD students from the Distributed Systems research group proposed a prizewinning plan 1 to make the Bernoulliborg building sustainable in terms of waste disposal and the use of water and energy Winning the Green Mind Award 2 means that the proposed plan is being executed in the course of the year 2013 This thesis is part of the execution of the plan s energy savings solution that aims to save at least 7 7 107 514 kWh annually through automatically controlli
20. UR EU TS 31 Implementatio N a a a a aa asea a e A 32 4 1 Office EVI Meda aa 32 4T Kitchen Mr beet d e PR de d ined ade dps edt deas 33 4 1 2 M eetingatea i COR 34 4 1 3 Office ceo AA e EE eo d e Pa ts ap dl 34 4 14 HallWay nn 1 mere ete A Pa T en Rd HA CREER Ge aides 34 4 2 Architecture aote a eite cade gana da e e c Lp gta Ea a nds 34 42 1 Physical layers usce ter tet ter e er eee terea e nn een 34 4 2 2 Se vice layer cs aie aute ela RR ERO LR UE a UH Ra c cde s 35 4 2 3 Persistence layer aite aa ea dau arb ea uio Ra d db une 35 4 2 4 09ic layer AAA 35 4 3 CoMmponentseu u int en einen teh tainted 35 431 OrientDB cod nr te Sahin bis ad 36 4 3 2 Sensor and actuator gateway concierne 36 4 3 3 Consumption measurement ssssssssssseeeeeeenenennnneneenenen nnne eiie 37 4 3 4 Controller ita re cree ep re die anid a pe Pd n ed feri ede 38 4 3 5 Orchestrator cnin ee be pl eco eta dre dne ra cen b a d i pa e e d b d C nip ps 40 4 3 6 700K66 GF 556 d nee Ha HI ete inten Laer t detto oe Pe LET ere ahi lene Re et as 40 4 3 Other components e a a vcre t et ne eR c due d fei ia Seka 41 AA COMMUNICATION iia 42 4 5 DISCUSSIO Masini 43 Theoretical results sisi lesen etc t tae reta te tica ed tfr eam e phe 45 5 1 Consumption Measures d D ile pen e HL Reden a 45 RENS A 45 5 1 2 PubliEareasa 48 5 2 Estimated energy saving potential and other researc
21. and also the guarantees of two graph databases can both be satisfactory for the final approach In case of the scalability of the database OrientDB clearly wins as it supports both read and write scalability The availability of both OrientDB and Neo4j are satisfactory even though Neo4j only supports high availability in the 27 73 enterprise edition In terms of performance is OrientDB a clear winner while Neo4j has the largest community largest usage and by far the best documentation in terms of quantity quality and diversity On the other hand Neo4j also has the most expensive enterprise edition and is missing important functionality for scalability in the free version The fact that the performance and scalability of OrientDB are better and that the documentation usage and community of OrientDB are also sufficient led to the decision to use OrientDB instead of Neo4j for the final approach 3 4 Message queue The message queue can be used by the components of the system to communicate with each other with or without knowing the physical location of the other components M essage queues can significantly simplify the implementation of the separate components and also improve performance scalability and reliability Another advantage of using a message queue is that the sender and receiver do not need to communicate with the message queue at the same time as the messages are stored onto the queue until the recipient receives the message As a
22. and 228 workdays per year the annual savings could amount to 153 9 kWh to 227 81 kWh per office annually 46 73 M easured between 15 220 watt Workstation 150 watt un o 0 T T T T T T T T T T T 01234567 8 9 1011 12 13 14 15 16 17 18 19 20 21 22 23 Hour of day T T T T T E Common consumption pattern of a workstation in the course of a day o WS WS2 1500 1000 4 E en DD 1 2 3 4 5 6 7 Day of week Mon Sun Common weekly consumption pattern for an office workstation Notes Researchers sometimes use their workstation to run experiments Common workstations still use 7 10 watt when turned off Workstations are measured through a power strip that contains the computer and monitor s This power strip might occasionally contain extra office appliances Figure 5 2 Energy consumption summary of workstations Workstation energy consumption is more irregular because of staff having different needs Some workstations contain better hardware in order to run research projects some staff members prefer dual monitors and some staff member use their laptop sometimes with an external monitor This means that the energy consumption ranges from around 15 watt for a small laptop to 220 watt for a dual monitor research workstation The top graph in Figure 5 2 shows the hourly consumption of a standard university issued computer which peaks a
23. anthr f Innen FIH63AP an 65 12 amp eur visited 25 06 2013 http www my knx shop net Eltako Funk Helligkeitssensor anthr f Innen FIH63AP an ADAFRUIT Flora Lux Sensor TSL2561 Light Sensor v1 0 ID 1246 7 95 Adafruit Industries Unique amp fun DIY electronics and kits visited 25 06 2013 http www adafruit com products 1246 NODNA Devantech SRFO8 Ultrasonic Range Finder noDNA Robotshop visited 26 06 2013 http nodna de Devantech SRF08 Ultrasonic Range Finder NODNA Devantech USB to I2C Interface module V2 noDNA Robotshop visited 26 06 2013 http nodna de USB to I2C Interface module ELECTROSHOP Eltako FTK raam deurcontact visited 26 06 2013 http www electroshop nl products Eltako FTK raam9 67B47967Ddeurcontact html AKKTOR STM 250 Magnetkontakt Funkmodul Akktor Shop f r Haus und Geb udeautomation visited 26 06 2013 http shop akktor de STM 250 M agnetkontakt Funkmodul LBL Standby Power Summary Table visited 28 06 2013 http standby Ibl gov summary table html ENECO Stroomprijzen bekijk de prijs per kWh Eneco visited 15 07 2013 http thuis eneco nl groene energie energieprijzen stroomprijzen W3 SPARQL Query Language for RDF visited 30 04 2013 http www w3 org TR rdf spargl query NEOTECHNOLOGY Neo Technology Neo4j World s Leading Graph Database Graph Database News visited 30 04 2013 http www neotechnology com NEOTECHNOLOCY Price List Neo Technology Neo4j G
24. bug fixes every month 59 Great documentation available such as guides tutorials videos and books 60 Improvements to vertical scalability are scheduled for 2013 55 Disadvantages High performance relies heavily on the available RAM of the system 56 Thehigh availability clustering features used for availability and horizontal scalability require the enterprise edition 57 Write scalability is limited Clients can write to slave nodes but the master node remains the limited factor 58 3 3 3 OrientDB OrientDB is an open source database with features from both document and graph database systems It is developed and supported by Orient Technologies and two of the three versions can be used free of charge for both commercial and free usage Professional support plans are available for 2 500 to 4 000 EUR depending on the license and length 64 There are three different versions of OrientDB released 49 50 Standard edition Apache2 license Support is limited to the community but professional support plans are available Price 0 EUR per year Graph edition Apache2 license includes TinkerPop Stack 65 Support is limited to the community but professional support plans are available Price 0 EUR per year Enterprise edition commercial license includes TinkerPop Stack Query Profiler Configurable Alerts and Production Support for 1 year This edition comes with professional email support that is
25. during the lunch hours it would mean that our above average consumption of the fridge during office hours could be eliminated On average the daily energy consumption of the fridge is around 0 4 kWh watt the energy consumption during the weekend in the lower graph of Figure 5 5 while during regular weekdays this amounts to 0 465 kWh As a result there is a difference of 0 065 kWh between a regular weekday and a day in the weekend This would mean a saving of 0 325 kWh 5 days 0 065 kWh per workweek and a yearly saving of 14 82 kWh 228 working days 0 065 kWh per fridge which amount to around a potential energy saving of 10 50 73 M easured between 560 755 watt when on Corridor Corridor2 Corridor 3 800 gt So 600 Corridor E 400 lighting u 0 T T T T T T T T 01234 5 6 7 8 9 101112 13 14 15 16 17 15 19 20 21 22 23 Hour of day Common consumption pattern combined corridor light fixtures in the course of one day 5Corridor1 Corridor2 Corridor 3 15000 10000 3 5000 o 1 2 3 4 5 6 7 Day of week Mon Sun Common weekly consumption pattern of combined corridor light fixtures in an office Notes Corridors are operated at building level and are only turned off completely including the Plugwise device at night and in the weekend Light are turned on around 7 30 and remain on till around 20 00 12 50 hours Alway
26. functionalities of the system WiFi RFID and NFC are all standards to establish communication between devices It was also shown in Chapter 3 that custom built sensors and actuators are often cheaper than their retail alternatives By exploring and selecting the correct sensors and actuators for a whole building the benefit cost ratio of the System can be maximized Due to time constraints the subject of optimizing the type location and number of sensors and actuators was only briefly examined One of the questions that arose was the need for a light sensor in order to control the lighting better But such light sensors are expensive Are light sensors best placed at a window per desk per light fixture or simply at the center of the room Another good extension would be to make corridor lighting motion activated but how many sensors would be needed with which range and where should the sensors be placed Questions about cost benefit ratio and placement arise for most of the sensors and actuators which justifies extensive research into this part of the system specifically Closely related to the choice of sensors is research into the range and network latency of larger indoor wireless sensor or actuator networks The area of sensors and actuators clearly provides room for a lot of research into the combinations and placement of different sensors and actuators and their overall usefulness Perhaps the most direct extension of this thesis is by the me
27. in graphs It is an open source database that is supported and developed by Neo Technology 48 There are three different versions of Neo4j released for different purposes 49 Community edition GPL license Only community support and is allowed only for open source projects Price 0 EUR per year Advanced edition commercial or AGPL license includes Advanced Monitoring This edition comes with professional email support that is available 10 hours a day 5 days a week Both open source projects and commercial projects are allowed Price 0 EUR per year for open source or 6 000 EUR per year for commercial projects Enterprise edition commercial or AGPL license includes High Availability Clustering Online Backups and Advanced Monitoring This edition comes with professional email support that is available 24 hours a day and 7 days a week Both open source and commercial projects are allowed Price 0 EUR per year for open source or 24 000 EUR per year for commercial projects Advantages m u I Support for many languages and frameworks 50 Active community for deployment development and problems 51 Used in production by large corporations such as Adobe Mozilla Lufthansa and Cisco 52 Excellent availability and horizontal read scaling characteristics for up to dozens of servers 54 Active development A new major version comes out approximately every six months with minor versions containing mostly
28. in the future The storage can also be used by components to store their state in so that a component can always continue from the previous state if necessary e g in the case of a hardware failure The main data stored in OrientDB is the energy consumption measured by the consumption component the variables created by the context the rules from the controller the user information from the dashboard and the state of the connected Plugwise devices from the SAGW Each component also registers 16 73 itself to the ZooKeeper instance so components can find each other if needed and also determine if all components are online 3 1 4 Logic layer The components belonging to the logic level are the high level components These components use all the available data and functionality from the lower layers in order to make calculated decisions about the state of devices This is also the layer where the system management components are located The controller is the component responsible for providing the system with the ability to automate the sensor and actuator behavior which can possibly influence the energy consumption The main function of the dashboard and mobile app is to raise awareness of the energy consumption to the employees 3 2 Sensors and actuators The sensors and actuators are necessary to provide energy consumption tracking and to potentially reduce the energy consumption These sensors and actuators need to create enough energy sav
29. level suggesting that the workstation is running but not actively used in the weekend In some cases it could be sufficient for staff to be able indicate time slots where the workstation should be active For example in the case of a web server that is only accessed during work hours it would be sufficient to leave that computer running all day even when the computer client indicates that the user is idle but is put it in standby at night and in the weekend It is hard to pinpoint the amount of potential energy saving because of the amount of different hardware being used University workstations still consume energy when they are turned off but a staff laptop that is unplugged will not When taking into account that a regular workstation still uses 7 to 10 watt per hour when turned off it would mean that in a 40 hour workweek 7 10 watt is being consumed 128 hours per week per workstation for nothing This comes down to approximately 1 kWh 2296 per workstation per week that can be saved by turning off the workstation completely at night and in the weekend 5 1 2 Public areas Public devices are located in the areas adjacent to the offices This includes the light fixtures in the corridor and the appliances located in the kitchen area M easured between 2 6 510 watt Microwave Microwave 300 watt e o 0 T T T T T 012345678 91011121314151617 Hour of day T 18 19 20 212223 Common consump
30. of Groningen in The Netherlands This consumption data is used to estimate the possible annual energy saving potential of the system of when it would have actually been operational Due to two unfinished components that were co developed by two other groups it was not possible to automatically influence the energy consumption with the Green M ind system In summary this thesis offers the following contributions Encourages numerous new research projects for bachelor and master thesis and allow the design and development of new homegrown ideas Provides accessibility to a live testing environment of actively used offices for the university Thedesigned system isto be used in production in the future rather than being another prototype Introduces a new system design for component based automation to achieve potential energy savings Provides energy consumption data from offices located at the University of Groningen in The Netherlands 1 4 Overview This section presents how the remainder of the thesis is organized Chapter 2 is about related work and introduces the theoretical framework and previous related research Chapter 3 provides an overview of the tools and methods being used where both the used hardware and software technologies are outlined Chapter 4 presents the architecture of the developed system as a whole The used solutions are explained while referring to the goals set in this introduction and their results are d
31. result the communication between different components is asynchronous The communication between the components is further discussed in Section 4 4 whereas the communication directions between the different components are depicted in Figure 3 1 Because of the aforementioned advantages the decision was made to use message queues for the communication of the system 3 4 1 Inventory An inventory of message queues are described in this subsection Each message queue has advantages and disadvantages that will be used as the input for the eventual approach The most described message queue choices are 132 133 134 135 136 ActiveM Q 124 Zer0M Q 125 Kafka 126 RabbitM Q 127 Qpid 128 Amazon SQS 129 HornetQ 130 Apollo 131 SSS SSS A A The following message brokers were dropped for consideration after an initial review Zer0MQ due to having no guaranteed delivery of messages 137 and that most functionality requires you to implement and combine various components of Zer0M Q 138 Apollo due to an inadequate documentation 131 and community 149 150 Amazon SQS due to being software as a service SaaS only 139 and having a pay as you go pricing scheme 140 Qpid due to number of bug reports 141 and having an inadequate community 142 143 ActiveMQ due to the number of user complaints 145 146 147 and having a low performance 135 136 144 Thethree remaining message queues are su
32. the system in Section 3 2 Finally an approach for the database is given in Section 3 3 followed by the message broker in Section 3 4 3 1 Design overview The Green Mind system has been built out of loosely coupled components that each have their own designated tasks This section briefly describes each component that is depicted in Figure 3 1 The components communicate with each other asynchronously through the Kafka message queue In Chapter 4 these components will be explained in more detail The architecture of the system is briefly addressed by this section in order of the layers bottom up that are shown in Figure 3 1 Logic Controller OrientDB Zookeeper d Service Context Orchestrator Physical PC client Plugwise Kafka unidirectional Kafka bidirectional OrientDB bidirectional qu gt Zookeeper bidirectional Figure 3 1 Architectural overview of the Green Mind system 15 73 The typical operation cycle is as follows the SAGW and SMS send messages to the context based on the changed values from their sensors and actuators the context processes these messages to create new variables and sends the variables to the controller the controller looks up the rules and executes them based on the received variables messages that require action taken are then sent from the controller to actuator the orchestrator then sends commands to
33. the effort required of installing Such a system in an existing or new building Based on previous research discussions with building managers and observations during walkarounds in the Bernoulliborg the potentials of saving energy are high and the system should be able to save energy even if the scale and volatile environment lower the efficiency of the system The hypothesis is thus as follows Installing a building automation system that controls multiple occupied offices has the potential to save energy 1 2 2 Measurable goals A total of three goals have been set in order to judge the effectiveness of the final implementation The details of these goals are described in this subsection and will be used in the discussion of the estimated results and the conclusion of this thesis later on 10 73 Goal 1 Lighting Based on the goal set by the proposed plan for the Green Mind Award the energy consumption of lighting should be lowered by 25 Lighting is currently motion enabled or always on which is not very efficient Goal 2 Electronic devices Based on the goal set by the proposed plan for the Green M ind Award the energy consumption of electronic devices should be lowered by 25 Non essential devices sometimes remain enabled unnecessarily e Q during absence night or weekend Goal 3 Return on investment Based on the goal set by the proposed plan for the Green Mind Award costs of installation should be returned within
34. the layer where the system management components are located Usually components in the logic layer have or would benefit from a full fletched graphical user interface as opposed to the lower level components where a console application suffices Even though the input is coming from the lower layers to the logic layer the processed data in the logic layer i e output will be used as input for the lower layers in turn 4 3 Components This section describes the function of the individual components from the different layers of the Green M ind system All the components are loosely coupled which means that components can operate independently of each other The added advantage is that the other components can remain running if another component fails 35 73 or wait until the failed component is online again Loose coupling also means that components find each other at runtime as opposed to static compile time binding The components communicate through a common message topic to make it easier to replace components or to run multiple instances in the future e g if a single component becomes a bottleneck Communication between the components is done with a Kafka message queue which is accessible by the components through ZooKeeper The communication is further discussed in Section 4 4 4 3 1 OrientDB After careful consideration see Section 3 2 OrientDB was picked as the main storage facilitator for the whole Green M ind system For some
35. users available to ask and answer questions For each of the seven guarantees there is one graph database to provide the 1 best solution with the other two graph databases providing the 2 or 3 best solution The guarantees are compared based on the closer review from subsections 3 3 2 3 3 3 and 3 3 4 and are shown in Table 3 1 A denotes an exceptional property a denotes a positive property whereas a denotes a negative property for a specific graph database 26 73 Scalability Availability Documentatio n Price 1 Read and write Internet network support 3 M aster slave replication Master failure leaves system non operational gna Similar performance as Neo4j 75 76 77 82 but severely lower than OrientDB gna Great documentation available 3 Smallest but active community 3 Always requires a license unless it is only for testing or research purposes High availability requires a separate configuration gu Only research projects Read and write Only local network support 1 M ulti master replication 1 Best performance 75 76 77 115 gna Great documentation available gn Substantial and active community 1 Nearly all features are available for free The enterprise edition is the cheapest compared to DEX and Neo4j gna Used by medium sized companies gna 3 Read 4 Internet network support Q
36. visited 25 07 2013 http wiki secondlife com wiki Message Queue Evaluation Notes Apache QPID 2FRed Hat MR G STACKOVERFLOW Feedback about ActiveM Q How about it Is it worth Stack Overflow visited 25 07 2013 http stackoverflow com questions 1809554 feedback about activemg how about it is it worth 6561777 6561777 NABBLE Qpid Apache Qpid users Mailing List Archive visited 25 07 2013 http apid 2158936 n2 nabble com Apache Qpid users f2158936 html NABBLE ActiveM Q Search for message apollo NOT subject svn OR commit OR jira OR build visited 25 07 2013 http activemq 2283324 n4 nabble com template NamlServlet jtp macro search page amp node 234 1804 amp query2message 63Aapollo4NOT 4subject963A9628svn XOR 4commit4OR4jira4OR4build9629 amp da ys 30 STACKOVERFLOW Questions containing apollo apache Stack Overflow visited 25 07 2013 http stackoverflow com search g apollo apache RABBITM Q RabbitM Q Commercial Services visited 26 07 2013 http www rabbitma com services html LINKEDIN Open 32dMHy3 Yt Ft of Xt SRiy oid pdeR Y S t 3S lj d dS Ch PO ELA SRly 30 visited 26 07 2013 http blog linkedin com 2011 01 11 open source linkedin kafka LINKEDIN Intra cluster Replication in Apache Kafka LinkedIn Engineering visited 26 07 2013 http engineering linkedin com kafka intra cluster replication apache kafka JBOSS HornetQ General FAQs HornetQ JBoss Community visited 29 07 2013 htt
37. 23 Hour of day Common consumption pattern of the fridge in the course of one day Fridge 600 3 200 0 TT 5 E i TY r a 1 2 3 4 5 6 7 Day of week Mon Sun Common weekly consumption pattern of the fridge Notes Consumes 16 6 watt on average when left undisturbed Figure 5 5 Energy consumption summary of the fridge in the kitchen The fridge in the kitchen is used to keep drinks and the lunches of the staff cooled The daily consumption shown in the top graph of Figure 5 5 shows that the fridge is cooled on intervals Energy consumption rises during offices hour when staff opens the fridge and thus more cooling is required Energy consumption peaks at lunch and dinner time The fridge cannot be turned off completely because the drinks and food in the fridge need to be cooled to stay fresh Previous research by the Distributed Systems of the University of Groningen has shown that the fridge is capable of cooling sufficiently when only turned on for 15 minutes of every hour 22 During these 15 minutes the fridge is cooling at the maximum energy usage Taking into account that the rated consumption of the fridge is 70 watt the consumption of the fridge during those 15 minutes per hour should be around 17 watt at most This is nearly the same as our measured average of 0 166 kWh when the fridge is left undisturbed As previous research in 22 also revealed that the cooling was sufficient even
38. 39 140 141 142 143 144 145 oa 146 147 148 149 150 151 152 153 154 155 156 157 158 ZEROM Q QM Q The Guide M Q The Guide visited 25 07 2013 http zquide zeromg org page 3aall AMAZON AWS Amazon Simple Queue Service SQS Queue Messaging Service visited 25 07 2013 http aws amazon com sas highlights AMAZON AWS Amazon Simple Queue Service SQS Queue Messaging Service visited 25 07 2013 http aws amazon com sqs pricing APACHE Issue Navigator ASF JIRA visited 25 07 2013 https issues apache org jira issues jql project 20in 20 QPID 2C 20PROTON 20AND 20issu etype 20 3D 20Bug 20AN D 20status 20 3D 200pen 20order 20by 20updatedDate 20d esc STACKOVERFLOW Questions containing qpid Stack Overflow visited 25 07 2013 http stackoverflow com search q gpid X AEON A quick message queue benchmark ActiveM Q RabbitM Q HornetQ QPID Apollo M uriel s Tech Blog visited 25 07 2013 http x aeon com wp 2013 04 10 a quick message queue benchmark activemg rabbitmg hornetg gpid apollo HIRAM CHIRINO visited 25 07 2013 http hiramchirino com stomp benchmark ec2 cl xlarge index html throughput to topic YCOM BINATOR ActiveM Q Not ready for prime time Hacker News visited 25 07 2013 https news ycombinator com item id 2588072 SECONDLIFE M essage Queue Evaluation Notes Second Life Wiki
39. 980461574999551012 posts default 8662850803828497527 HYPERGRAPHDB HypergraphDB A Graph Database visited 18 07 2013 http www hypergraphdb org blog entry http ww w blogger com feeds 1980461574999551012 posts default 173304913295485825 FRANZ AllegroGraph Editions visited 18 07 2013 http www franz com agraph allegrograph ag commercial edition Ihtml FRANZ Franz Inc Technical Support visited 18 07 2013 http www franz com support Ihtml STACKOVERFLOW Questions containing allegrograph Stack Overflow visited 18 07 2013 http stackoverflow com search q allegrograph FRANZ AllegroGraph RDFStore Web 3 0 s Database visited 18 07 2013 http www franz com agraph allegrograph OBJECTIVITY Objectivity TAnfiniteGraph visited 18 07 2013 http www objectivity com infinitegraph GOOGLE 60 InfiniteGraph Developer Forum Google Groups visited 18 07 2013 https groups google com forum forum infinitegraph GOOGLE regarding data replication and failover in IG 3 0 Google Groups visited 22 07 2013 https groups google com forum topic infinitegraph yfdU dLkK6eA SLIDESHARE OrientDB distributed architecture 1 1 visited 22 07 2013 http www slideshare net lvca orientdb distributed architecture 11 GOOGLE Horizontal scalability in orientdb Google Groups visited 22 07 2013 https groups google com forum topic orient database tF W 1KdDTOM A GITHUB Home orientechnologies orientdb Wiki GitHub
40. VN Index of commons proper jexl tags visited 15 08 2013 http svn apache org viewvc commons proper jexl tags CODEHAUS M VEL Home visited 15 08 2013 http mvel codehaus org CODEHAUS M VEL Language Guide for 2 0 visited 15 08 2013 http mvel codehaus org Language4Guide4for42 0 CBS CBS StatLine Aardgas en elektriciteit gemiddelde prijzen van eindverbruikers visited 25 08 2013 http statline cbs nl StatW eb publication DM SLNL amp PA 81309NED amp D1 0 1 5 8 13 15 amp D2 0 amp D3 1 amp D4 19 24 29 amp HDR T amp STB G2 G3 G1 amp VW T PLUGWISE Home Extension Plugwise visited 25 08 2013 http www plugwise com idpluatype f home home extension KNM I KNMI Klimaatdata en advies Informatie over verleden weer visited 25 08 2013 http www knmi nl klimatologie maand en seizoensoverzichten jaar jaar12 html GEM IDDELDGEZIEN Gemiddelde aantal werkdagen Gemiddeldgezien nl visited 25 08 2013 http gemiddeldgezien nl meer gemiddelden 106 gemiddelde aantal werkdagen Eric Bush J rg Nipkow Barbara Josephy Susanne Heutling and Rainer Griesshammer Strategies to enhance energy efficiency of coffee machines In EEDAL Conference paper ID vol 75 2009 A D Galasiu G R Newsham C Suvagau D M Sander Energy saving lighting control systems for open plan offices a field study Leukos v 4 no 1 July 2007 pp 7 29 Sheng Yuan Yang A novel cloud information agent system with Web service t
41. a cluster 181 Clients can write to each node 181 Uses a variant of master slave for high availability scalability 181 Larger community compared to Kafka 3 167 168 169 170 171 172 Disadvantages The amount of configuration required for high availability scalability 158 Large but unreadable and complex documentation e g often no explanations lack of structured online examples 160 Lowest performance compared to Kafka 1 and RabbitM Q 27 135 136 143 144 175 Not used by large companies except for Last fm 179 Requires shared storage for scalability that is accessible by all nodes 181 3 4 5 Overview The following guarantees must be provided by the message queue Scalability It is likely that one message queue will be used per building university or company With the amount of messages sent e g sensor measurements it is important for a message queue to handle increased load in such a way that the throughput can increase when resources e g more servers are added Availability As the components of the system depend on the data that is sent through the message queue it is important that the message queue is always accessible Without the data important decisions cannot be taken to influence the energy consumption Performance If the message queue is unable to cope with a large amount of data sent through the message queue i e the data is coming in at a higher rate
42. ages High availability scalability is built in and easy to setup 157 The client maintains what was received and can request previous messages e g at startup if needed 159 The setup and code required is the easiest compared to RabbitM Q 2 and HornetQ 3 163 164 165 Well written and complete documentation 161 162 Used by large companies such as LinkedIn Twitter Tumblr Foursquare Netflix and M ozilla 166 Best performance compared to RabbitM Q 25 and HornetQ 3 135 136 143 144 175 Uses a variant of master slave for high availability scalability 153 159 Clients can write to each node 159 Replication support for a cluster 159 SS a I Disadvantages Smallest community compared to RabbitM Q 1 and HornetQ 27 167 168 169 170 171 172 Requires external application i e Zookeeper for discovering message topics and nodes 159 No professional support available 3 4 4 HornetQ HornetQ is an open source multi protocol and asynchronous messaging system designed with usability in mind HornetQ was originally developed as JBoss M essaging 2 0 and is currently developed by the community with assistance from a small team located at Red Hat JBoss 154 Red Hat JBoss also provides professional support for HornetQ through their regular subscription model 155 29 73 Advantages SE m I Professional support available by two companies 155 Replication support for
43. al 7 3 Testing the system sensors and actuators Although the system components were independently tested testing the system components together uncovered several problems with the system namely the unfinished SAGW and context components Due to these unfinished components the remaining part of the system could only be tested for communication and basic functionality By finishing the SAGW and context components the system as a whole could be tested not only for communication and functionality but also for other aspects such as the stability and latency e g between input from the SAGW and output from the orchestrator or between other components This advanced testing helps to obtain a better understanding of the possibilities and limitations of each component and the system as a whole In this case it would be important to investigate the factors that affect the working of the system Currently the system used for testing is used in production In an environment specifically designed for the purpose of testing one could make and test all necessary changes to the system without affecting the production environment 7 4 Security It becomes necessary in networked environments to secure the system from unauthorized access as an unwanted visitor can disrupt the activities of the system Even read only access may be dangerous as sensitive information about the internal structure can be derived from the exchanged messages In these cases it is
44. alability can persist the messages that are sent allows dients to receive pending messages after a failure 184 Kafka easily integrates with ZooKeeper and also allows the SAGW to easily communicate to multiple components which is especially useful when other components also need to be notified when the state of a device changes e g the context component can be notified instantly when the orchestrator component changes the state of a device to eliminate the need for polling the state of devices continuously 4 3 3 Consumption measurement Plugwise devices are used to measure the consumption of devices or to turn connected device on or off Each Plugwise device has three different buffers that contain consumption data 1 A 1 second buffer that contains the energy consumption of the last second 2 A8 seconds buffer that contains the energy consumption of the past 8 seconds 3 Ahistory buffer that stores the energy consumption per hour for more than a year An older and smaller research project at the University of Groningen 188 retrieved the one and eight seconds buffers at a fixed interval but in the larger office environment this turned out to be impractical due to the extra processing time needed for the larger amount of Plugwise devices The latency of the Plugwise Zigbee network also increased due to the increased spread of the Plugwise devices relative to other Plugwise devices and the architectural obstructions between the Plugwise de
45. and Windows Phone could be of assistance too e g for instant notifications to the administrators 7 6 Extending rule support for controller The rules for the controller provide the ability to control the connected sensors and actuators Even though rules can be specified for a whole range of similar devices e g lights computers it is possible that there are multiple similar rules due to personal preferences Perhaps the rules should be designed in such a way that users can specify their own preferences for each rule For example by adding a generic rule for lights where each user or room can specify the minimum amount of light intensity before the associated lights are switched off Another area for further exploration is the possibility of overriding a rule or device by the user e g to keep a computer turned on during the weekend Both the personal preferences for rules and the ability to override the rules can prove to be useful for the user acceptance rate of the system 62 73 References 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 T A NGUYEN F NIZAM IC Bernoulliborg The building of sustainability November 2012 wW DpgiK AA t o2d zdat Ayr OAABYY yA SNA 2FDMy 38 yp sited 25 06 2013 http www rug nl about us who are we sustainability green mind award RUG Energy lt Rijksuniversiteit Groningen visited 02 08 2013 http ww
46. ans of the sensors and actuators 7 2 Comfort enhancement In terms of functionality the system focuses on monitoring and reducing the energy consumption As a result it also affects the comfort of the user both positively e g no need to manually turn off the computer after working hours and negatively e g after a short break the computer is put to sleep A possible reason for the negative impact on the comfort is due to limited means of the system to accurately detect occupancy and location Perhaps it is possible to limit the discomfort or even increase the overall comfort through the use of more or more precise sensors and actuators e g automatically turning on the computer when the user enters the office This could lead to a better user acceptance rate of the system Potential pitfalls are the impact of the implementation on the user s everyday routine For example making users carry along an tag as described in Section 7 1 might solve the localization and occupancy sensing but might provide additional discomfort because the user might have to swipe the tag against a control surface every time when entering or leaving a room A perk of the Green M ind system is that it does not require the user to do anything differently than without the system The Plugwise devices and computer client do not 60 73 require user interaction no learning curve and the system does not drastically influence the everyday flow when it is operation
47. app component is larger and has more diverse tasks than most of the other components It handles both the authentication for remote viewers and also serves all the different presentational formats to the users on the different device types e g public display in the building applications on mobile phones Even though an employee can view the energy consumption of his own workstation for a public display in the building it is not desired to single out a specific user With the data model used in the database for the energy consumption i e a user is located in a room and that room in turn belongs to a floor of a building it is 41 73 possible to change the detail of the data For public displays the dashboard will use a more general representation such as energy consumption on a per floor or per building basis 4 4 Communication All the components in the Green Mind system communicate with each other through the use of JSON objects JSON was picked because of the good readability simple syntax and ease of use JSON is also less verbose than alternatives like XM L which decreases the size of the message payloads Developers of components that communicate with each other dictate the content of the messages that are communicated with each other Depending on the data that the component requires the body of the JSON object may change For the sake of traceability with asynchronous communication and readability and consistency in generic messa
48. area does not contain any equipment except for a projector The following types of devices can be controlled for the meeting area 1xlighting fixture 1xoutlet strip for portable devices 4 1 3 Office Offices are the most common areas and differ in size shape interior and occupancy throughout the building Despite the different design the general usage remains the same and thus the following can be controlled inside an office Ncomputer and appliances where N depends on the amount of workplaces and people in an office 1xlighting fixture It is worth noting that the systems of the university and the portable devices of the employees run different operating systems This is not only important in the case of putting computers in standby but also for the implementation of the Green M ind system itself because the system has to run on all the systems of the university 4 1 4 Hallway Hallways are the common entrance halls that run between the different areas The hallways allow entrance to the offices kitchen area and meeting area that are controlled by the system In total there are two hallways being controlled which are depicted in Figures 4 1 and 4 2 The following types of devices can be controlled for each hallway 1xlighting fixture 4 2 Architecture The Green Mind system has been built out of loosely coupled components that each have their own designated tasks These components need to work on the systems of the u
49. articular entity in the building such as an office device or person and relies on a Bluetooth based indoor positioning system that requires software to be installed on the mobile phone of each user From the three studies that focus on all three subsystems two of these studies 27 28 solely focus on houses instead of our focus on offices The other study namely the GreenerBuildings project 23 considers multiple buildings such as offices universities and hotels for deployment The architectural model of the 13 73 GreenerBuildings project was used as a starting point for the architectural model for the work in this thesis Unlike GreenerBuildings this thesis solely focuses on offices does not interact with the Smart Grid and is deployed in actively used offices instead of a lab building 23 There is one study in particular that precedes the work in this thesis Georgievski et al 22 proposed a system to monitor and control the office environment while coupling it with the Smart Grid The work in this thesis differs in the fact that it does not connect with the Smart Grid and is destined to be used for a whole building in production rather than an isolated test site of three offices 14 73 Chapter 3 Technical approach In this chapter the hardware and software technologies of the system are outlined This chapter starts with a design overview of the system in Section 3 1 followed by an approach for the sensors and actuators of
50. ation of the sensors and actuators and higher level components should not be bothered with this conversion Instead of the high level components communicating with a specific component of each specific sensor which is likely to change in the future they 36 73 communicate with the SAGW only The main advantage is that only a new interface has to be added to the SAGW when a new sensor is added making sensors a lot easier to manage For the Plugwise devices a completely new implementation was written in Java to be able to turn devices on and off as well as retrieving the consumption information for all the available devices The sensors or actuators for the SAGW are also not limited to hardware only In the case of the sleep management server SMS component it was necessary to run a small piece of software on the workstation computers to determine if a computer was being used active or idle state or to put the computer into another state on sleep hibernate and off The abilities of the software installed on the workstation computers are handled by the SAGW just like any other actuator Components communicate with the SAGW through a Kafka message queue and JavaScript Object Notation JSON objects JSON is text based and designed for human readable data interchange REST would have been a valid alternative for communication between the components but a message queue has several important advantages compared to REST e g allows for better sc
51. available 9 hours a day 7 days a week Price 4 000 EUR per year 24 73 Advantages m S Supports both document and graph data models 68 Uses query language similar to SQL 61 Support for many languages and frameworks 112 Distributed architecture horizontal scalability read and write to improve performance and availability 106 107 Replication can be configured to use either synchronous mode always consistent or asynchronous mode eventual consistency 106 Supports multi master replication allowing each node to read and write to the database 106 Great documentation available such as guides tutorials videos and books 107 Improvements to vertical scalability e g distributed transactions are scheduled for version 2 0 which is expected to be released before the end of 2013 79 Active development a new major version comes out approximately every six months with minor versions containing mostly bug fixes every month 110 111 Disadvantages Not used in production by large corporations only in medium sized corporations 67 Scalability replication is limited to local networks only 69 73 78 Community is not as large and active as Neo4j but questions are usually answered within a day or two 73 3 3 4 DEX DEX is a high performance and scalable graph database system It is developed supported and licensed under a proprietary license by Sparsity Technol
52. aving higher components access them directly e g through the use of shared software libraries or other common code for the individual sensor interfaces it becomes easier to maintain and update the code of the interfaces without significant chances ending up with different kinds of implementations per component or that fixes are not being shared with developers of other components Another reason to use a physical layer is that devices exist that can only be accessed through one physical location e g a USB stick is required to access the connected devices or accessed by one component at a time e g a serial port connection that cannot be shared The complexity of the diversity of sensors and actuators is also something that the developers of other components should not have to deal with 4 2 2 Service layer Components in the service layer use the functionality of the physical layer to provide added value for the higher layers The service components provide functionality that exceeds the get and set level of the physical layer and simplifies sensor and actuator access even further Some components create new output themselves e g when processing and converting the consumption data from the Plugwise devices while others just process high level commands and translate those to commands that can be used by the SAGW 4 2 3 Persistence layer The persistence layer acts as a system wide long term storage The data is stored by the components in s
53. azon Simple Queue Service SQS Queue M essaging Service visited 25 07 2013 http aws amazon com sqs JBOSS HornetQ putting the buzz in messaging JBoss Community visited 25 07 2013 http www jboss org hornetq APACHE Apollo visited 25 07 2013 http activemq apache org apollo STACKOVERFLOW message queue ActiveM Q or RabbitM Q or ZeroM Q or Stack Overflow visited 25 07 2013 http stackoverflow com questions 731233 activemq or rabbitmq or zero mg or SECONDLIFE M essage Queue Evaluation Notes Second Life Wiki visited 25 07 2013 http wiki secondlife com wiki M essage Queue Evaluation Notes PREDIC8 ActiveM Q Qpid HornetQ and RabbitM Q in Comparison visited 25 07 2013 http www predic8 com activemg hornetg rabbitmg apollo qpid comparison htm JAVAPLEX High Performance Message Queues get Reviewed Java Plex Exploring Java Technologies visited 25 07 2013 http www javaplex com blog high performance message queues get reviewed X AEON A quick message queue benchmark ActiveM Q RabbitM Q HornetQ QPID Apollo M uriel s Tech Blog visited 25 07 2013 http x aeon com wp 2013 04 10 a quick message queue benchmark activemg rabbitmg hornetq qpid apollo ZEROM Q zeromq dev is OM Q guaranteed in order delivery visited 25 07 2013 http lists zeromq org pipermail zeromq dev 2009 M ay 000808 html 68 73 138 1
54. base are the rules of the controller component the variables of the context component the sensor list that is maintained by the SAGW and user management of the dashboard amp mobile app component 4 3 2 Sensor and actuator gateway Sensors and actuators are an important part of the Green M ind system Without the sensors there is no data for components to base their decisions upon and without actuators it is not possible to influence the energy consumption A big concern with the sensors and actuators is that they return different types of values and that the method of retrieving those values differs per vendor On top of that do the sensors and actuators often not easily interact with software and require extra effort is to easily interact with them e g through Java Because the Green Mind system is currently a research framework in particular with the initial implementation running at the University of Groningen it is almost certain that different hardware will be added in the future The SAGW is designed to generalize and simplify access to sensors and actuators for the other high level components Low level communication such as serial communication with the Plugwise devices is implemented in different interfaces which the SAGW uses to execute the commands and requests from the other components The SAGW also converts any encoded values such as hexadecimal representations to useful values because conversion often depends on the implement
55. bject to a closer review RabbitM Q is described in Subsection 3 4 2 Kafka in Subsection 3 4 3 and HornetQ in Subsection 3 4 4 28 73 3 4 2 RabbitM Q RabbitM Q is an open source enterprise messaging queue based on the emerging AMQP 183 standard RabbitM Q is developed and supported by GoPivotal Inc whereas VM Ware provides a commercial licensed release of RabbitM Q under the vFabric brand name 151 Professional support plans are available by both GoPivotal Inc and VM Ware for undisclosed prices Advantages Largest community compared to Kafka 3 and HornetQ 2 167 168 169 170 171 172 Professional support available by two companies Well written and complete documentation 173 174 Used by large companies such as VM Ware 176 Nokia 176 Huffington Post 176 Aadhaar 176 Digg 178 and NASA 178 Replication support for a cluster 180 Uses master slave for high availability scalability 156 SS m m Disadvantages Theamount of configuration required for high availability scalability 156 Clients can only write to the master node 132 156 182 3 4 3 Kafka Kafka is an open source high throughput publish subscribe messaging queue that is scalable durable and fault tolerant by design Kafka was originally developed at LinkedIn 152 before it became an Apache project Linkedin is still involved in the development of Kafka and uses Kafka internally for several processes 153 Advant
56. by large sized Used by almost no large companies companies companies Table 3 2 Overview between RabbitM Q Kafka and HornetQ about scalability and availability performance documentation community and usage 3 4 6 Conclusion HornetQ never provides the best solution for any of the six guarantees as shown in Table 3 2 While providing the second best solution for usage community scalability and availability it also provides the worst solution for performance and documentation As RabbitMQ and Kafka both provide better or equal solutions for all six guarantees the decision was made to not consider HornetQ anymore for the final approach The remaining two memory queues RabbitMQ and Kafka provide an equal amount of best solutions RabbitMQ provides the best solution for documentation community and shared usage whereas Kafka provides the best solution for scalability and availability performance and shared usage However not all guarantees have the same importance and there are also guarantees that are satisfactory for the final approach for both remaining memory queues The documentation of RabbitM Q is better and the community larger but the documentation and community of Kafka are satisfactory too In terms of the performance scalability and availability is Kafka the clear winner The fact the scalability and availability of Kafka are better than that of RabbitM Q and that the performance of Kafka is at least two times that o
57. components is communication not enough and the data they create is not only important at the moment of creation but also for later reference e g for research or presentational purposes This historical data needs to be stored in a central place so that other components can also access this common data without the help of communication from other components By relying heavily on other components the chance of losing data increases because when there are more components there are also more possible points of failure By providing a storage system especially with data replication such as OrientDB to increase reliability and availability it is not always necessary to go through the different components to gather the required data A good example of important data is the indirect cooperation between the consumption component and the dashboard amp mobile app component The dashboard amp mobile app component uses the gathered data from the consumption component to present the user with graphs and other statistics about the energy consumption Instead of having the consumption component communicate all the new consumption measurements to the dashboard amp mobile app component directly which would then also need to store it somewhere to show historical data the consumption measurements are stored in the database directly The dashboard amp mobile app then retrieves this data from the database on request Other data that is also stored in the data
58. current functionality of the Green Mind system These potential energy savings are also compared to previous work in Section 5 2 QN ar gy S LNI 2TBj 3u LNI K o5 YE XSR 6 nex tK Qa 2T y ARAORAEE t 2 6 RSOKS 28 BAK GH HOt 2NJy3 SSS A Gydfit of 40 hours 8 hours per day and 5 days in the week totaling to 228 workdays 204 per year after holidays This leaves 128 hours per week 24 hours a day and 7 days a week minus 40 working hours that are outside office hours The Bernoulliborg building is operated from 7 30 to 20 00 which is based on the measured time that the corridor lighting is turned on Public area equipment e g coffee machine microwave is expected to be online for the total 12 30 hours which leaves a remaining 11 5 hours per day that the building is considered empty For a whole week this results in 105 5 hours 11 5 hours per day for 5 days per week plus 24 hours per day for 2 weekend days that are outside office hours 5 1 Consumption measurements Over a course of six weeks from mid May to late July the office environment on the fifth floor of the Bernoulliborg has been monitored 24 hours a day and 7 days a week M easurements are divided into two segments The first segment consists out of the devices located in the offices These devices and their consumption depends on occupancy of the offices and the usage of the devices by the staff that is located in that specific office The second segment consists ou
59. d Science CloudCom CLOUDCOM 12 IEEE Computer Society Washington DC USA 363 370 SPARSITY TECHNOLOGIES Sparsity technologies DEX high performance graph database visited 22 07 2013 http www sparsity technologies com dex_tutorials5 name HighAvailability NEO4J Neo4j the Graph Database Learn Develop Participate visited 23 07 2013 http www neo4j org GITHUB Titan Distributed Graph Database visited 23 07 2013 http thinkaurelius github io titan HYPERGRAPHDB HypergraphDB A Graph Database visited 23 07 2013 http www hypergraphdb org index FRANZ AllegroGraph RDFStore Web 3 0 s Database visited 23 07 2013 http www franz com agraph allegrograph SPARSITY TECHNOLOGIES Sparsity technologies DEX high performance graph database visited 23 07 2013 http www sparsity technologies com dex php OBJECTIVITY Objectivity TAnfiniteGraph visited 23 07 2013 http www objectivity com infinitegraph ORIENTDB OrientDB Graph Document NoSQL dbms visited 23 07 2013 http www orientdb org 1t1 1 9p L OS Gba v N Index visited 25 07 2013 http activemq apache org ZEROM Q The Intelligent Transport Layer zeromg visited 25 07 2013 http www zeromg org APACHE Apache Kafka visited 25 07 2013 http kafka apache org RABBITM Q RabbitM Q Messaging that just works visited 25 07 2013 http www rabbitmg com APACHE ApacK v LRVIDOSABR 44 07 2013 http qpid apache org AM AZON AWS Am
60. dies 2 800000 messagesminute how nokias here uses rabbitma to make real time traffic maps VMWARE RabbitM Q VM ware vFabric Blog VM ware Blogs visited 01 08 2013 http blogs vmware com vfabric rabbitmg RABBITM Q RabbitM Q How To visited 01 08 2013 http www rabbitmg com how html specific examples BLOGSPOT The HornetQ Team Blog visited 01 08 2013 http hornetg blogspot nl RABBITM Q RabbitM Q Highly Available Queues visited 01 08 2013 http www rabbitmg com ha html SOURCEFORGE Chapter 39 High Availability and Failover visited 01 08 2013 http hornetq sourceforge net docs hornetq 2 0 0 GA user manual en html ha html DIRECTI RabbitM Q vs Apache ActiveM Q vs Apache qpid visited 02 08 2013 http bhavin directi com rabbitmg vs apache activemg vs apache qpid AMQP Home AMQP visited 25 08 2013 http www amqgp org STACKOVERFLOW M essage Queue vs Web Services Stack Overflow visited 07 08 2013 http stackoverflow com questions 2383912 message queue vs w eb services 70 73 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 GITHUB device plugwise perl lib Device Plugwise pm at releases hollie device plugwise perl GitHub visited 07 08 2013 https github com hollie device plugwise perl blob releases lib Device Plugwise pm Maar
61. e Special thanks go to our second supervisor Alexandru Telea for his friendly advice motivation and insightful comments We would also like to thank Faris Nizamic and Tuan Anh Nguyen for their encouragement and invaluable suggestions We have greatly benefited from the discussions with our fellow team members Mattijs Meiboom Rosario Contarino Amedeo M erlo Brian Setz and Ruurtjan Pul Our deepest appreciation goes to the volunteers that allowed us to install sensors and monitor their offices Without their support this thesis would not have been possible Table of contents ADS A 2 Acknowledgments an RA E tton gren teet 3 Table of contents iia annie die iaa 4 HISEOTEIgUEGS Reese ren ta 7 EE 8 Introduetion PR 9 1l CoOnEeXE eem e edad AS EC Re E D UE Cr RR dete era 9 1 2 Research goals oc oed ted dt oe ie o E pet eerie t eer d ado 10 1 2 I Hypothesis A ter EE M RO RF SA RR T ce etd tes 10 1 2 2 Measurable goals act station anat bei dba tu adeb segre 10 1 2 3 Planned solution 2t ntc a ratae a Dat d aa Eget 11 1 2 4 Bpect ti h u a etae ret e e tete a ea I een vr feet 11 1 3 CONCIDUTIONS a a s eaa ea ata dv pa i GR ha s a n cct e has 12 1 4 OVeLVIeW c cocer ta rela te aua inel ee eui a feu fa T pla dedu 12 Related WOrKii ebbe o bio taeda gud geo bai anat dependen b d aao Belle 13 Technical approaeh aote erts tete das Gael tes
62. e API that is usually described in the context of HTTP whereas message queues are separate software components that can be used for asynchronous communication where messages are placed onto a queue until the recipient receives them The data is stored in a central database and components find each other through ZooKeeper an open source server which enables highly reliable distributed coordination It is envisioned that over time new sensors and actuators are added to the system as well as new components that provide better or completely new functionality so that the system keeps updating and improving The system is named the Green M ind system which refers to its origin as a proposal for the Green M ind Award 1 2 4 Expectation Through the loose coupling of the individual components the system should scale better more easily adapt to new components and be easier to manage than a conventional BM S By using a dynamic combination of small hardware that integrates with the existing building design and dedicated software the system has the potential to save energy at a relatively low cost without the loss of comfort and even the potential to increase comfort Based on previous local projects and preliminary research the initial design of the system should be able to reach the goals of this thesis while also forming the basic framework for future research 11 73 1 3 Contributions The main driver behind this project is the focus of the universi
63. e energy consumption of multiple devices over the course of six weeks The next period of six weeks might differ because of all kinds of influences e g temperate sunlight holidays Even during these six weeks it became dear that the usage and thus the energy consumption can be irregular at times This is especially so in the case of the energy consumption of regular workstations when compared to workstations that are being used for research Even with the irregularities the measurements and estimates are still applicable By using averages and taking into account the excesses the estimated potential energy savings give a credible and applicable scenario that could be reached through building automation There is no denying that the measurements presented in Chapter 5 show that there is still unnecessary energy consumed by all the measured devices Especially with the motion sensors already being used by office lighting it shows that more automation is necessary to increase the efficiency even further 5 3 2 Lighting The goal of lowering the consumption of lighting by 25 can easily be reached when turning lights off when there is enough natural light available The estimations show that at least 36 can be saved A potential problem with realizing these savings is determining the right amount of light sensors and placing them correctly in order to optimize the savings The unique design of the blue facade of the building which uses an externa
64. e of the fifth floor Map of the offices and meeting area on the north wing of the fifth floor Content of a hour entry as returned by the SAGW Example of a simplified rule to turn a computer off when it is idle and it is after 20 00 Example of an extended rule to turn a light on when the weather is cloudy JSON object of a typical request JSON object of a typical successful response JSON object of a typical error response Energy consumption summary of office lights Energy consumption summary of workstations Energy consumption summary of the kitchen microwave oven Energy consumption summary of coffee machine in the kitchen Energy consumption summary of the fridge in the kitchen Energy consumption summary of lights in the corridor Energy consumption summary of the electric water boiler in the kitchen Energy consumption summary of the meeting room 15 32 33 38 39 39 42 42 43 46 47 48 49 50 51 52 53 List of Tables Table 3 1 Table 3 2 Table 4 1 Table 5 1 Overview between Neo4j DEX and OrientDB about scalability availability performance documentation community price and usage Overview between RabbitMQ Kafka and HornetQ about scalability and availability performance documentation community and usage The three available values for the status field in a response JSON object Summary of the potential energy saving found through the measurements 27 3l 43 54 Chapter 1
65. e on off sleep or hibernate in case of the workstations for the SMS whereas the possible statuses of a variable depend on the type of the device for the SAGW 4 3 6 ZooKeeper Not all the components of the Green Mind system are running on the same computer One reason can be because of hardware limitations and another reason is the importance of the geographical location e g making sure that the person responsible for maintenance is as close to the system as possible for direct access Each component registers the IP addresses and ports used with ZooKeeper at startup ZooKeeper allows components to find each other without the need to know the IP address of each other which reduces the amount of configuration necessary for each component ZooKeeper also makes it easy for a component to find out if all components are up and running because a component does not have a ZooKeeper entry anymore after it fails Integration with ZooKeeper greatly simplifies maintenance and automates discovery of the other components Another advantage is that Kafka uses ZooKeeper out of the box as well for the same purpose As Kafka also has the ability to connect to a remote ZooKeeper installation instead of using the shipped installation only one installation is required for both Kafka and the components of the Green M ind system 40 73 4 3 7 Other components Because of the scale of the system some of the components i e context SMS and dashboard were impl
66. e research to regular workstation ratio is lower and for the fifth wing of the Bernoulliborg it should definitely be encouraged to simply not run each research workstation 24 hours a day and 7 days a week just out of convenience 56 73 5 3 4 Appliances The measured appliances in the kitchen area can be divided into two categories first there are the microwave 7 estimated potential energy savings and the fridge 9 estimated potential energy savings Both QYoX R SAE AS dk e Brent yydbfiep KKK 14 Mcr 2y AA Sy 2F he SS N BNBZIK N Plugwise devices The potential savings do not justify the costs of two individual Plugwise devices for the fridge and the microwave These potential savings are also hard to increase through further automation because the fridge simply needs to cool its contents and the microwave only has an idle consumption of 2 6 watt The annual consumption of the microwave is not extremely high when compared to the other measurements despite the high consumption of a microwave when active because of the short time span necessary to heat its contents In the case of the microwave a possible solution would be to use a single Plugwise device for both the microwave and the coffee machine because they have a similar usage pattern This leaves the second category with the coffee machine and the electric water boiler which have estimated Lpibyiik CSySNe d Oys 2 Tyr yR nor Nj PK ETK Y 2y 0 N yds 2F S S Y t OOyS 6 Xt 0
67. echniques Example of an energy saving multi agent system Expert Systems with Applications Volume 40 Issue 5 April 2013 Pages 1758 1785 ISSN 0957 4174 Sunghoi Park M yeong in Choi Byeongkwan Kang Sehyun Park Design and Implementation of Smart Energy M anagement System for Reducing Power Consumption Using ZigBee Wireless 71 73 Communication M odule Procedia Computer Science Volume 19 2013 Pages 662 668 ISSN 1877 0509 72173 Glossary AcronymExplanation ACL Access Control List AMQP Advanced M essage Queuing Protocol API Application Programming Interface BMS Building Management System GUID Globally Unique Identifier HVAC Heating Ventilation and Air Conditioning JSON JavaScript Object Notation NFC Near Field Communication PIR Passive InfraRed RDF Resource Description Framework REST Representational State Transfer RFID Radio Frequency Identification SaaS Software as a Service SAGW Sensor and Actuator GateWay SMS Sleep M anagement Server 73173
68. ectedwatt 330 66562 correctedkwh 0 33066562 Figure 4 3 Content of an hour entry as returned by the SAGW 4 3 4 Controller The controller is the sole component responsible for providing the system with the ability to automate sensor and actuator behavior which can positively affect the energy usage Without the controller the system is not able to act upon the sensor and actuator variables but can only monitor and store the variables from the sensors and actuators By adding a component to the system that can analyze the sensor variables and control the actuators it can read and change the state of the connected device e g turn off a monitor in the weekend and therefore achieve potential energy savings Because of existing experience in the Distributed Systems group the desire to dynamically add new commands without the need to recompile and the variations of user preferences e g some users like their offices dark the decisions was made to develop a rule based controller that uses a format that is easy to read and write These rules provide the ability for the system to act upon the sensor and actuator variables e g in the weekend all computers should be turned off or be in hibernate 38 73 Because the rules are to be parsed evaluated and executed the decision was made to use an expression language to facilitate the querying and manipulation of Java objects at runtime M VEL 199 was chosen after careful consideration b
69. either the SMS or SAGW depending on the received message the SMS or SAGW then executes the requested action e g turn off the Plugwise device the cycle repeats The components do not wait until a full cycle is completed e g a change in the sensor value can occur at any moment 3 1 1 Physical layer All the different kinds of sensors and actuators are located in the physical layer A sensor measures a quantity e g light movement that can be read by an external component whereas an actuator performs actions based on commands from an external component e g turn on the power put a computer to hibernate These sensors and actuators can be accessed through the sensor and actuator gateway SAGW which provides custom interfaces to all the sensors and actuators so that the system can easily access the functionality of the actuators and sensors no matter the vendor or version of the hardware Components in the physical layer only gather information from the sensors and actuators and merely allow the higher layers to easily access the sensors and actuators The SAGW has currently only support for two sensors and actuators namely the PC client and the Plugwise interface Other sensors and actuators can be added when needed The PC client is responsible for detecting activity on the computer and changing the state of the computer i e on off hibernate or sleep The Plugwise interface is responsible for easy access to the Plugwise devices tha
70. emented by other teams The tasks of these components in the Green Mind system are discussed in this subsection Context Sensors and actuators are mostly limited to one simple task e g a PIR sensor is either true or false when there is movement A common problem in the Bernoulliborg building is that the lights in the offices also contain such a PIR sensor which means that if an employee is working quietly at a desk the sensor will not detect any movement and will turn off the lights The task of the context is to avoid such problems by creating a completely new variable by combining values from multiple sources In the case of the lights this could mean that a new variable is created called RoomOccupancy which is set according to the activity readings from both the computer client and the PIR sensor This means that when the PIR sensor does not register movement and the computer is still being used the lights will remain on The variables generated by the context expand the possibilities of the system even further by creating valuable additional information from the sensor input The context either requests or receives a notification about the changes of the sensor and actuator states through the Kafka message queue These states are then processed by Twitter Storm to form the new context variables through a predefined truth table The variables both processed and original are stored in the OrientDB database and are also passed on to the controller
71. ensor the energy consumption sensor can be used to switch off the lights when the intensity measured is greater than a predefined value The measured intensity from the light sensor can also be used to switch on the lights using the energy consumption sensor when the value is below the predefined value On the other hand the PIR sensor can be used for locations such as rooms and hallways where lights should be switched on when movement is detected As PIR sensors only detect constant movement a delay is often required to have the lights not constantly switch between on and off Multiple possibilities of switching lights on or off are achievable in case both a light sensor and PIR sensor combined with an energy consumption sensor are used To maximize the potential energy savings in this case it is recommended to let the PIR sensor have priority over the light sensor when the light intensity is under a predefined value and to let the light sensor have priority over the PIR sensor when the light intensity is above a predefined value In general the key elements to decide the sensors for a location are the amount of users that pass through the area the amount of daylight entering the area the type of area and the personal preferences of the users This resulted in the decision to have the ability to use all three possible scenarios 3 2 4 Other There are other devices e g a beamer a warm water boiler and an oven microwave that are not available
72. estimate potential savings for lighting to be 2296 to 4296 depending on the level of automation utilized The proposed plan of energy saving through light sensors reached savings of 11 to 47 in a field study conducted in open plan offices 206 A large multi campus building experiment by Sheng Yuan Yang 207 with an energy saving multi agent system also managed to reach savings up to 20 for lighting 5 2 2 Workstation Based on the measured energy consumption of workstations that had been turned off it should be possible to save 2296 annually per regular workstation by cutting the power to these workstations completely at night and during weekends For workstations it is only possible to estimate the potential energy savings for the remaining energy consumption when the workstation is supposed to be turned off completely This is caused because there is no data about idle time available yet from the Green M ind system Other research that contained details about workstation energy savings mostly only focus on computers and incorporate multiple stages of energy savings such as disabling monitors putting the computer into sleep mode and shutting down computers outside office hours A study in Japan 8 showed that an average desktop computer consumed around 3096 of its energy during idling and 4096 in non business hours On the other 54 73 hand there was the study conducted by Park et al 208 that only showed an energy saving of 0 2 by sett
73. etween several other expression language ANTLR 189 which always requires a separate executable 194 and is mostly developed for tree based language recognition 189 Jep 190 which is only available as a trial version 195 JUEL 195 which has no active development 197 and low performance compared to M VEL 192 OGNL 191 which has low performance compared to M VEL 192 JEXL 193 which has no active development 198 and low performance compared to M VEL 192 SS m a MVEL provides the controller with the ability to evaluate the rules with nearly the same performance as regular Java code and also provides a list of powerful features suitable for parsing evaluating and executing the rules e g projections folds property navigation 200 The variables that the controller received from the context can be used in the rules A variable contains a type e g computer an identifier e g 127 0 0 1 a name e g is idle and a value e g false A list of all variables and their state is initially send to the controller by the context the moment the controller is started whereas only the changes of the variables are send by the context afterwards Variables from the same type are grouped together e g computers for easy access by the rules The controller itself provides one additional variable that is not received from the context namely a date variable The date provides the rules with the ability to use the day
74. f RabbitM Q 175 led to the decision to use Kafka instead of RabbitM Q for the final approach 31 73 Chapter 4 Implementation Chapter 4 describes the office environment and the architecture of the Green M ind system that was installed to monitor and influence the energy consumption First in Section 4 1 the layout of the offices is introduced in order to give a better understanding of the operating environment and the scale Then an overview is given of the complete architecture in Section 4 2 before discussing the individual components in detail in Section 4 3 Finally Section 4 4 handles the communication between the components and how the components find each other 4 1 Office environment The main area to be controlled is located on the top fifth floor of the Bernoulliborg building and consists of two adjacent corridors that connect in a right angle at the north east corner of the building A In total twelve areas are controlled by the system located on the fifth floor and connected to each other through the two included hallways The areas that are controlled by the system are as follows 8xoffice 2xhallway 1xkitchen area 1xmeeting area The first eight offices pertain to employees from the Distributed Systems group at the university After the system is tested and found stable more offices from the university are going to be connected the Green Mind system IE Eh 8 Il O HHHH
75. ges a basic template was specified to which all the messages of components should adhere This basic template is depicted in Figure 4 6 id 7FBOA415 7AED 440F A4D7 EDFEB299ED2C data method requestdevicestate parameters mac 000D6F000098CO09E Figure 4 6 JSON object of a typical request Each request that a component sends should have an id by which the message can be uniquely identified In the case of the request in Figure 4 6 the id consists of a regular Globally Unique Identifier GUID which is a unique reference number that can be used as an identifier The id field can be used to couple request and response which is especially useful in cases where communication happens asynchronous As a message queue is always asynchronous the id field is mandatory to trace the response back to a request request 7FB0A415 7AED 440F A4D7 EDFEB299ED2C Status success data mac 000D6F000098CO09E state on Figure 4 7 JSON object of a typical successful response When a component responds to a request it reuses the original id from the request and places it into the request field of the response object There is also a status field that indicates whether the component was able to correctly process the request or not Figure 4 7 and Figure 4 8 depict an example response message with a successful and an error status respectively 42 73 request 7FB0A415 7AED 440F A4D7 EDFEB299ED2C
76. gs project website visited 25 08 2013 http www greenerbuildings eu approach Tuan Anh Nguyen Marco Aiello Energy intelligent buildings based on user activity A survey Energy and Buildings Volume 56 January 2013 Pages 244 257 AC iDorm visited 25 08 2013 http cswww essex ac uk iieg idorm htm Paul Davidsson and M agnus Boman Distributed M onitoring and Control of Office Buildings by Embedded Agents Information Sciences Vol 171 4 Elsevier 2005 COLORADO M ichael C M ozer visited 25 08 2013 http www cs colorado edu mozer index php dir Research Projects Adaptive9620house L Hawarah S Ploix M Jacomino User behavior prediction in energy consumption in housing using Bayesian networks Proceedings of the 10th international conference on Artificial intelligence and soft computing Part I ICAISC 10 Springer Verlag Berlin Heidelberg 2010 pp 3727379 ARDUINO Shield Xbee w o RF module A000021 Bert j o Njd y2 2N community and electronics visited 25 06 2013 http store arduino cc eu index php main page product info amp cPath 11 5 amp products id ARDUINO Arduino Uno Rev3 A000066 B4 3j o Njdiy2 N community and electronics visited 25 06 2013 http store arduino cc eu index php main page product info amp cPath 11 12 amp products id 195 SPARKFUN Flexiforce Pressure Sensor 100lbs SparkFun Electronics visited 25 06 2013 https www sparkfun com products 8685 a hbblte M
77. h sss 53 ZA PE A A A 54 5 2 2 Workstation nl ettet A A a 54 5 2 3 Appliances ene an be el e eii ia eb ede POR ee riti ies 55 5 3 DISCUSS iria a i d bat ee Ban 55 5 3 1 Reliability ard applicability z iti at tec ern 55 SEAT RN 55 5 3 3 WOFPKSEaLlOTIS seo te A umet be m end Eee E c CR Gee vni dr A repe TU S 56 5 3 4 Appliances oe end ee De rre pu ha cert ut t D RU E dte pot ur rdc Que ee nee 57 Concl sion ETE 58 Future work c ta er patet de p e had d e e Pte ga patate d tpe eti ra 60 7 1 Extending and improving the sensors and actuators sse 60 7 2 Comfort enhancement fece eee A 60 7 3 Testing the system sensors and actuators conciernen 61 y Eoi um E ai 61 7 5 Administrator panel et cis ce erret ee eere rrt ed sede Ee Een PP EEUU EUER eae erede a eed Lado vs 61 7 6 Extending rule support for controller nennen nnne nnns 62 References usi meme da eid epa ie de da ede inde eb o Ped de aep Hd edu o ead ia 63 GloSsaty eate teta sande hd dein Hee granted ta P e a tage E na 73 List of Figures Figure 1 1 Figure 3 1 Figure 4 1 Figure 4 2 Figure 4 3 Figure 4 4 Figure 4 5 Figure 4 6 Figure 4 7 Figure 4 8 Figure 5 1 Figure 5 2 Figure 5 3 Figure 5 4 Figure 5 5 Figure 5 6 Figure 5 7 Figure 5 8 The Bernoulliborg building Architectural overview of the Green M ind system Map of the offices and kitchen area on the east sid
78. hanges are received from the context The flexibility of the rules provide the Green Mind system with the ability to achieve potential energy savings by acting upon the sensor and actuator variables that are send by the context The output from the controller is sent to the orchestrator which in turn will send a message to the SAGW or the SM S based on the input of the controller 4 3 5 Orchestrator In order to keep the controller as simple as possible the execution of the plans made by the controller is left to the orchestrator The orchestrator component acts a buffer between the physical layer and the controller and takes care of sending the commands at the correct time as well as the order of execution Each plan can be started and stopped at any time or paused and resumed if necessary To limit the chance of the orchestrator component becoming a bottleneck the orchestrator is multithreaded and implemented in such a way that multiple instances of the orchestrator can execute simultaneously The orchestrator communicates with the SMS and SAGW depending on the instructions received from the controller The SMS handles the requests from the orchestrator to change the state of workstations and the SAGW handles the state change requests for the other devices The current state change requests are straightforward and consist out of a unique identifier for the device e g a MAC address IP address and the desired status of a variable The status can b
79. hat it is possible to only turn on specific fixtures and partially light the hallway For example in the corner where the hallway in the 51 73 east side of the building meets the hallway of the north wing there are no windows to let in sunlight and some additional lighting might be appreciated by some staff members whereas the rest of the hallways does not need any additional lighting M easured between 43 577 watt when active Waterboiler TT 600 Electric water 400 boiler E 200 0 ne T E T TE T T T Y T T E T e if 1 01234 5 6 7 8 91011121314151617181920212223 Hour of day Common consumption pattern of the water boiler in the course of one day Waterboiler 2000 1500 4 g 1000 500 0 Z T T Y T gt E 1 1 2 3 4 5 6 7 Day of week Mon Sun Common weekly consumption pattern of the electric water boiler Notes Does not consume any power when not heating Consumes 20 watt per hour on average when left undisturbed for a whole day Figure 5 7 Energy consumption summary of the electric water boiler in the kitchen There is also an electric water boiler installed in the kitchen area Figure 5 7 shows the consumption information that was measured for this boiler The top graphs shows that the boiler uses power intermittently to keep its content heated This is useful during office hours when there is need for insta
80. he potential energy savings are only 15 In the specific cases of specialized research workstations tailor made rules have to be designed in consultation with the responsible researchers in order to fully optimize the energy saving potential Costs of the implementation are limited to that of the hardware such as new sensors and actuators For this thesis the costs are created by the purchase of the Plugwise devices The third goal of costs being returned within at most seven years is reached mainly because of the annual energy savings per light fixture of an S N HR Brot KS 2342 FS d Jy S N 5 2NJ d Gey SAK t dia X6 By amp Nfs not justified in every case but one has to take into account that the installation of multiple sensors does allow the localization of excessive energy use By experimenting with temporary sensor and actuator placement and creating different groupings according to the readings the cost efficiency can be further improved The fruits of the thesis have provided the University of Groningen with numerous new possibilities for new research projects through accessibility to a live testing environment of actively used offices A new modular system design has been introduced that allows for future incremental enhancements but is also able to operate outside a controlled experimental environment The currently designed functionality limited as it may be is already capable of saving energy within the set goals Through its loosel
81. hether lights should be switched on or off The light intensity is important to determine context and to influence the energy consumption The following sensors can be applied to measure movement energy consumption and light intensity Light sensor To measure the light intensity in lux near the location of the sensor Energy consumption sensor To monitor the energy consumption of the lights either over a single light or multiple lights at the same time 21 73 PIR sensor To determine if a room or hallway is occupied or not To influence the energy consumption of the lights there is three possible scenario l Axenergy consumption sensor Bx light sensor and Cx optional power strip where the values of A B and C depend on the situation of the lights 2 Axenergy consumption sensor Bx light sensor Cx PIR sensor and Dx optional power strip where the values of A B Cand D depend on the situation of the lights 3 Axenergy consumption sensor Bx PIR sensor and Cx optional power strip where the values of A B and C depend on the situation of the lights The type and amount of sensors required per location depend on the situation For a busy hallway one light sensor and one energy consumption sensor with a power strip to connects all lights can suffice to maximize potential energy savings On another hand a hallway that is not often used can use a PIR sensor instead of a light sensor In case the situation requires a light s
82. ht Figure 5 4 Energy consumption summary of coffee machine in the kitchen Coffee machines are common in offices and Figure 5 4 shows that the idle consumption of such a device is high even when it is not being used Most measurements were 58 watt although roughly every six hours a measurement of about 38 watt occurs when the coffee machine was idle Thus on average the coffee machine uses 56 watt when there is no staff member to us it There is no need for the coffee machine to be active outside office hours which means a potential energy saving of 5900 watt 56 watt 105 5 hours per week which is around 50 savings of the total weekly energy consumption of the coffee machine Potential issues with turning off the coffee machine at night and in weekends is that the machine might not be capable of recovering from long downtime Besides the hardware limitations there might also be issues with the freshness of the products stored inside the machine when the power is lost over longer periods of time The fact remains that the monitored coffee machine consumes a substantial amount of energy at night and in the weekend Judging by the recommendations by Bush et al 205 regarding the energy efficiency of coffee machines it should be possible to disable the coffee machine completely 49 73 Se M easured between 3 6 291 watt Fridge 40 30 Fridge 8 20 10 0 4 012345 6 7 8 9 1011121314 15 16 17 18 19 20 21 22
83. hub com nuvolabase orientdb wiki Replication GITHUB Binary Data nuvolabase orientdb Wiki GitHub visited 01 05 2013 https github com nuvolabase orientdb wiki Binary Data GITHUB Distributed Configuration nuvolabase orientdb Wiki GitHub visited 01 05 2013 https github com nuvolabase orientdb wiki Distributed Configuration GOOGLE Scalability and performance Google Groups visited 01 05 2013 https groups google com forum msg orient database crIFYreBTdo ORftK5SAHFg4 GOOGLE Google Groups visited 01 05 2013 https groups google com forum forum orient database GOOGLE Performance of find references Google Groups visited 01 05 2013 https groups google com forum topic orient database 7VoiDE_5hLY GOOGLE OrientDB and neo4 Google Groups visited 01 05 2013 https groups google com forum fromgroups topic orient database Cbeokg6mFYw GITHUB GraphDB Comparison nuvolabase orientdb Wiki GitHub visited 01 05 2013 https github com nuvolabase orientdb wiki GraphDB Comparison GOOGLE DatabaseBenchmarks orient Orient Database Benchmarks NoSQL document database light portable and fast Supports ACID Tx Indexes asynch queries SQL layer clustering etc Google Project Hosting visited 01 05 2013 http code google com p orient wiki DatabaseBenchmarks SOFTPEDIA OrientDB 1 3 0 Changelog visited 01 05 2013 http mac softpedia com progChangelog Orient Changelog 75279 html GITHUB D
84. ible Judging that 3096 of the energy consumption is caused when the computer is idle 8 it should still be possible to reach the goal of 2596 energy savings especially when the research workstations get restrictions also It should be noted that the idle time is handled by the computer client which puts the computer into sleep mode and thus the corresponding monetary savings do not change anything about the cost of the Plugwise device being too high Although the workstations on their own do not reach the goal of saving 2596 of the energy consumption on devices the savings of the workstations combined with those of the kitchen reaches the goal From the workstations nine could be classified as regular workstations This would mean a total annual energy consumption of 2022 57 kWh 9 224 73 kWh with a total estimated potential annual saving of 457 47 kWh Combined with the measurements from the kitchen this would add up to 3381 7 kWh 1359 13 kWh for the kitchen see Subsection 5 3 4 and an estimated potential energy saving of 932 34 kWh totaling 27 596 on energy savings for all the devices combined When the research computers are also taken into account with a total of 2848 58 kWh annually then the potential energy savings drastically change for the worse With a total energy consumption of 6230 28 kWh annually the estimated potential savings for devices will only amount to 1596 and thus miss the initial goal of 25 On the whole building th
85. in high quantities but that are still very common in office environments and used very regularly throughout the day These devices can be turned off outside office hours to save power For these devices there is at least one variable to be measured to determine the context Energy consumption where the measurement for determining context can be done over each device A baseline measurement of the energy consumption compared to later measurements can measure how effective the solution is Variable X where X can be used to influence the energy consumption of a device The only requirement for the variable is that a sensor is required for the measurements The following sensors can be applied to measure movement energy consumption and light intensity Energy consumption sensor To monitor the energy consumption of the device SensorX To measure variable X Even though these devices are not placed in the computer and lighting category it is always possible to include other variables e g movement or sensors e g PIR sensor that can be used to influence the energy consumption Scenarios of using sensors other than the energy consumption sensor depend on the type of 22 73 device Most devices are only used during office hours and can be switched off safely otherwise whereas other devices can maximize their potential energy savings by using additional sensors e g using a PIR sensor to only enable the stereo system when moveme
86. ing the time of power usage or by using a motion sensor to control a laptop The multi campus building experiment by Sheng Yuan Yang 207 showed energy savings of 15 for computers 5 2 3 Appliances All the other measured devices are located in the public kitchen area something that is present in most offices but not explicitly mentioned in detail by other papers The fridge and microwave have an energy saving potential of 9 and 7 respectively The measurements in 208 also include a microwave and the measured energy savings for that microwave add up to 3 3 For the coffee machine a potential energy saving of 56 is estimated mainly because the high energy consumption of the machine outside office hours The electric water boiler has the potential to save 2896 of its energy consumption Both suffer from the flaw that they do not enter a standby mode and thus keep heating water even without demand No comparison material could be found for the energy consumption of the water boiler and coffee machine 5 3 Discussion In this section the measurements and estimations about potential energy savings from Chapter 5 are discussed in relation to the costs of the sensors and actuators and the usage of those sensors and actuators 5 3 1 Reliability and applicability Reliability and making predictions on energy consumption measurements is something that is always debatable The implementation of the Green Mind system has been able to reliably measure th
87. ings in order to justify their costs 3 2 1 Inventory An inventory of sensors and actuators that are feasible technically monetarily and privacy wise to integrate into the system are described in this subsection Each sensor and actuator has advantages and disadvantages that will be used as the input for the eventual approach Seat pressure sensor A seat pressure sensor can be placed at a location where it can detect the pressure of an occupant sitting on the seat This can either be inside below or on top of the seat As soon as an occupant sits on the seat the Seat pressure sensor detects the weight and signals it to the receiving party Advantages Detect room occupancy unless the seat is moved to another room Disadvantages Does not detect computer usage e g moving the seat to another desk for a meeting Notall seats are suitable for placing a seat pressure sensor e g wooden seats Requires significant effort and costs to install and integrate in existing environment Price 65 HiaB custom built excluding development and build costs 29 30 31 or prefabricated 32 33 Energy consumption sensor An energy consumption sensor can detect the energy consumption of connected appliances e g computer or lights The sensor can be placed between the appliance plug and the socket for easy installation The sensors can also be equipped with the ability to wirelessly switch on or off appliances making the sensors not on
88. ining part of the rule should be executed it is possible to improve the performance of such rules by determining which specific types and identifiers are used in the predicate For example it is not required to evaluate a predicate of a rule that only uses computer 10 0 0 1 when a variable change for computer 127 0 0 1 is received from the context All in all the simplified rules provide the users with an easy and fast way to act upon the sensor and actuator values to positively influence the energy consumption 39 73 The extended rules are direct Java code and provide the users to create rules that go beyond the possibilities of the simplified rules The extended rules do not feature a predicate and are always evaluated after a variable change is received from the context due to the sheer amount of expression possibilities However the ability to execute direct Java does impose a security risk that is not present with the simplified rules Depending on the need for more complex rules and the gains in energy saving potential of such rules the use of extended rules could be justified to provide more possibilities to increase the overall energy saving potential of the system Each time a new rule is added to the controller it is automatically stored to the database After the controller is restarted the rules are retrieved from the database and it is determined which rules use which variables in order to improve performance for when variable c
89. ion as very cheap occupancy sensors because most of the work being done in the offices involves a workstation one way or another These basic actuator and sensor capabilities combined with the knowledge of the time of day already makes for a system capable of saving energy in the Bernoulliborg building Workstations can be turned off at night and in the weekend Lights can be turned off during the day while normally the built in motion sensor of the office lights would still turn the lights on even if there is enough sunlight available The sensor capabilities of the Green M ind system are currently somewhat limited but the SAGW provides the possibility to add more sensors over time The current system would mostly benefit from a light illuminance sensor and a better means to detect occupancy The light illuminance sensor would allow the system to better manage the lighting in specific rooms according to the available sunlight instead of only relying on the time of day and occupancy A big drawback of such sensors is that commercial solutions are rigid and expensive The current viable occupancy sensor is cheap but has some drawbacks e g no location and amount of occupants and are all delay based that limit the potential energy savings The investment for all offices is large especially without proper research into the kind of sensors and their effectiveness e g what is the correct positioning of such sensors and which value ranges should be sup
90. iscussed Chapter 5 discusses the potential results and contains the collected energy consumption measurements The estimated results are compared to results of previous research and an explanation is given for why the measurements are as they are in Chapter 5 as well Finally Chapter 6 concludes this thesis with a recap of the research goals and a summary of the objectives and results followed by future work in Chapter 7 12 73 Chapter 2 Related work Numerous researches have analyzed and shown the energy saving potential in office buildings Lighting has been pointed out by various sources 10 11 12 13 as an area with significant energy improvement potentials Besides lighting office equipment e g computer and monitor influences the energy usage to a great extend Kawamoto et al 8 Table 4 and Webber et al 9 Table 3 show that a large percentage of computers and monitors are not turned off but in standby mode during non office hours 40 in Japan and 75 in US According to the authors of 8 computers and monitors in offices are found to be in standby mode for no less than 21 hours a day HVAC systems can also contribute to the possible energy and cost savings according to several studies 14 15 16 This thesis focuses only on lighting and office equipment for energy saving potential as the HVAC installed at the Bernoulliborg building is only manageable on a per wing basis rather than on the envisioned per room basis BM S sy
91. isk and memory resources 83 Active development a new major version comes out approximately every six months 112 25 73 Disadvantages Does not support many languages and frameworks 66 Inadequate community the number of new posts per week on the community forums are limited but questions are usually answered within a week 87 Expensive compared to Neo4j and OrientDB A failure of the master node leaves the system non operational despite a master slave implementation 89 Not used in production by large corporations the available scenarios only describe research projects with limited periods 88 There are no minor version releases 112 3 3 5 Overview The following guarantees must be provided by the graph database Scalability It is likely that one database will be used per building university or company With the amount of sensor measurements done it is important for a database to handle increased load in such a way that the throughput can increase when resources e g more servers are added Availability As the components of the system depend on the data that is stored in the database it is important that the database is always accessible Without the data important decisions cannot be taken to influence the energy consumption and is it possible that sensor measurements from a specific time are lost Performance If the database is unable to cope with a large amount of new sensor measurements i e
92. istributed transactions Issue 1418 nuvolabase orientdb GitHub visited 01 05 2013 https github com nuvolabase orientdb issues 1418 SPARSITY TECHNOLOGIES Sparsity technologies DEX high performance graph database visited 02 05 2013 http www sparsity technologies com dex licenses SPARSITY TECHNOLOGIES Sparsity technologies DEX high performance graph database visited 02 05 2013 http www sparsity technologies com dex_pricelist SPARSITY TECHNOLOGIES DEX Analytical Use Case Benchmark Wikipedia visited 02 05 2013 http sparsity technologies com blog p 228 SPARSITY TECHNOLOGIES Sparsity technologies DEX high performance graph database visited 02 05 2013 http www sparsity technologies com dex SPARSITY TECHNOLOGIES Sparsity technologies DEX high performance graph database visited 02 05 2013 http www sparsity technologies com dex tutorialsb name ndex SPARSITY TECHNOLOGIES Sparsity technologies DEX high performance graph database visited 02 05 2013 http www sparsity technologies com dex_ releases FACEBOOK Sparsity Technologies visited 02 05 2013 http www facebook com pages Sparsity Technologies 158957260788675 GOOGLE 17 DEXgdb Google Groups visited 02 05 2013 https groups google com forum fromgroups forum dexgdb SPARSITY TECHNOLOGIES Sparsity technologies DEX high performance graph database visited 02 05 2013 http www sparsity technologies com partners clients
93. l frame on top with partially printed transparent plates attached might limit the incoming daylight Light sensors and detailed occupancy models are necessary to form a better estimate Judging by the similar ranges 11 to 47 15 206 in energy savings from other papers and in particular the 20 energy savings on lighting in campus buildings 207 it is safe to conclude that the goal of 25 can be reached 55 73 kS Y 2y56 NY 4 Oys FN BRWYN KMB d uxLSNEIK TROIS c KSy 19 MR dy KS 32r 2 have a return of investment within seven years this would amount 2 KS LPA amp SoXAe 2 d HyR HNIdyR Bhia LEN fixture in sensor equipment In most cases it is sufficient to group multiple fixtures together For example the current installation groups two or three fixtures that are present in a single office together into one Plugwise S tek St Gal k AIBA XI 33 for the installation of one or more light sensors in the office 5 3 3 Workstations Research workstations consume considerably more energy than their regular counterparts and are mostly turned on 24 hours a day and 7 days a week The estimation in Table 5 1 of an energy savings potential of 2296 per regular workstation does not represent workstations that are used for research because they are continuously turned on When looking at the annual savings for all the workstations combined including the research workstations the potential energy savings are likely to turn out lower On the
94. ly useful to actively monitor but also to influence the energy consumption Advantages Insight in the energy consumption of connected appliances Switch appliances on or off e g after working hours Easy installation 17 73 wirelessly switch on or off appliances A power strip provides an easy and cheap way to increase the number of connected appliances e g for connecting multiple lights Disadvantages Protrudes from the socket in case the sensor is not integrated into the wall socket Easy for users to avoid i e by using another socket or by removing the energy consumption sensor Price 35 ymB 6 7 PIR sensor A passive infrared PIR sensor can be placed at a location where it can detect movement PIR sensors only detect movement rather than the location of that movement As the range and field of view of a PIR sensor is limited it is often placed at the ceiling in order to maximize coverage As soon as a PIR sensor detects movement it signals it to the receiving party Advantages Detect room occupancy Easy installation Unobtrusive placement Available as multi sensor light intensity and PIR Disadvantages Reliability varies e g by direct sunlight or direct wind from an air conditioner or heater Does not detect the location of movement Requires constant movement Does not detect very slow movement Price 60 HMj B custom built excluding development and build costs 29
95. mputer usage and influence the energy consumption of the computer there are two possible scenarios 1 1x ultrasonic range sensor 1x seat pressure sensor and 1x computer software sensor As single sensors the ultrasonic range sensor and seat pressure sensor both have severe shortcomings but reliable computer usage detection can be achieved by combining both sensors 2 1x computer software sensor The computer software sensor will measure the amount of time the pc is idle Based on the computer usage and particularly when the computer is not being used the computer s power status i e on S3 S4 or off can be changed to influence the energy consumption In both scenarios can the computer software sensor be used to put the computer in a different power status to achieve potential energy savings whereas the operating system will automatically put the monitor in standby mode when the power status of the computer changes In the second scenario the computer software sensor can detect computer usage e g by monitoring the input devices even though it can also be used to improve the accuracy of the first scenario The other two sensors of the first scenario both have their shortcomings but combining the sensors can achieve reliable computer usage detection The advantage of using these two sensors especially when combined with the computer software sensor is that computer usage can be detected almost instantly and also with better accuracy compa
96. n Master slave replication High availability requires the enterprise edition gna Similar performance as DEX 75 76 77 82 but severely lower than OrientDB 1 H Best documentation available 1 Largest and most active community gna Enterprise edition is the most expensive only for commercial projects High availability requires the enterprise edition 1 Used by large sized companies Table 3 1 Overview between Neo4j DEX and OrientDB about scalability availability performance documentation community price and usage 3 3 5 Conclusion As shown in Table 3 1 the DEX graph database only provides the best solution for scalability Even though scalability is absolutely important for the components of the system OrientDB provides almost similar scalability features with the only scalability limitation being no internet network support As the DEX graph database provides the 2 or 3 best solution for all the other guarantees it is not considered anymore for the final approach The remaining two graph databases Neo4j and OrientDB both have best solutions for different guarantees OrientDB excels in scalability availability performance and price whereas Neo4j excels in documentation and community However Table 3 1 also shows that the guarantees of Neo4j and OrientDB are intertwined with respect to each other Even though all guarantees are important some are more important than others
97. ng heating ventilation and air condition HVAC systems lights appliances and providing energy consumption tracking HVAC systems are applied to control the indoor environment comfort e g room temperature and ventilation The Bernoulliborg is located on the Zernike Complex in the Dutch city Groningen and was officially opened in February 2008 It accommodates 350 staff members and 500 students The overall size is 33 by 83 meters and a height of 27 meters lt houses both the Faculty of Mathematics and the Faculty Natural Sciences as well as the departments of mathematics computing sciences and artificial intelligence Figure 1 1 shows the facade of the building Common areas such as a library cafeteria and lecture rooms are located on the two lower levels Offices Figure L1 The Bernoulliborg building are located in three separate wings on the three upper floors of the building In order to reach the proposed minimal energy consumption saving of 7 796 a system has to be designed that can alter the state of devices such that energy can be saved without the loss of comfort or drastic alterations to the building s structure A building management system BM S is already present in the building which is accessible remotely by the campus management BM S systems are widely used in buildings to control and monitor the electrical equipment of the building such as lighting and HVAC systems However the BMS located at the Bernoulliborg buildi
98. ng is limited to controlling electrical equipment on a building level The only segmentation in the system is at most on a per wing basis which is nowhere near the envisioned granularity of management on a per person or per room level The control of the BMS is limited and not easily accessible by third party software it was thus not integrated into the design of the new system 9 73 Due to the limitations of the BMS it is necessary to install new sensors and actuators so that it becomes clear what or who uses the most energy and to measure how much exactly is being consumed at what time These sensors and actuators also use energy but need to create enough energy savings to justify their costs A live experiment was created to monitor eight actively used offices located on the fifth floor in the north wing the leftmost wing in the picture of Figure 1 in order to measure energy consumption test the possible sensors determine their usefulness and gain knowledge about the interaction of office workers Along with the offices the experiment also includes the adjoining hallway meeting room toilets and kitchen communal area The plan proposed for the Green M ind Award encompasses more than just energy saving and the system in this thesis thus takes into account the other components that will be integrated with the system over time 1 2 Research goals First in Subsection 1 2 1 the hypothesis for this thesis is provided In order to narr
99. niversity which consist out of different hardware and operating systems The majority of the components were therefore written in Java because an easy to understand cross platform programming language was preferred and Java was already commonly used in the Distributed Systems group at the university There are two exceptions the backend of the Dashboard amp Mobile App was developed in Python and the PC client was developed in C to eliminate the dependency to install Java on each system In Section 3 1 a design overview was briefly addressed about the components and basic architecture The components will be explained in detail in Section 4 3 but first the overall architecture of the system is addressed in the order of the layers bottom up that are shown in Figure 3 1 4 2 1 Physical layer All the different kinds of sensors and actuators are located in the physical layer These sensors can be accessed through the sensor and actuator gateway SAGW which provides custom interfaces to all the sensors so that the system can easily access the functionality of the actuators and sensors no matter the vendor or version of the hardware Components in the physical layer only gather information from the sensors and actuators They do not process the reading or send out commands by themselves the components merely allow the higher layers to easily access the sensors and actuators By allowing access to devices in the physical layer instead of 34 73 h
100. nt hot water but not so much at night and in the weekend when the building is empty When no hot water is tapped from the faucet the boiler consumes about 20 watt per hour on average per day This continuous consumption is clearly visible in the weekends where the consumption evens out at just below 500 watt per day as visible in the lower graph of Figure 5 7 Energy can be saved by disabling the water boiler outside office hours This would mean that 105 5 hours of consumption can be saved It might take time for the boiler to reheat its content completely so starting before working hours might be necessary This would still leave around 100 hours roughly an hour is required as depicted in Figure 5 7 which adds up to 2 kWh per week 23 of potential energy savings 52 73 M easured between 43 577 watt when used TN Thursdayx Thursday y Meeting room 300 z 200 _ 3 100 0 4 3 012345678 9 101112131415 16 17 18 19 20 21 22 23 Hour of day Common consumption pattern of the meeting room in the course of one day Weekx Weeky 1500 1000 4 3 500 0 1 2 3 4 5 6 7 Day of week Mon Sun Common weekly consumption pattern of the electric water boiler Notes Beamer is unplugged after usage No other appliances are present in the room Usage is sporadic Figure 5 8 Energy consumption summary of the meeting room The meeting r
101. nt is detected In all possible scenarios is an energy consumption sensor required to determine how effective the solution is 3 3 Databases The database is necessary to store and retrieve the data e g the sensor measurements for the components of the system The data can be stored in traditional relational databases using tables The primary difference is that in a graph database the relationships are stored at an individual record level whereas in a relational database the structure is defined at a higher level i e the table definitions One of the main advantages is that graph databases provides index free adjacency i e graph traversals can be performed with no index lookups that can lead to a much better performance Each sensor measurement contains relationships to other data such as the floor building time and date and room or user As a large amount i e 1 000 000 of sensor measurements with a traditional database caused severe performance issues e g retrieve the total energy consumption for a certain hour on a specific date the decision was made to use a graph database instead of a traditional database This also enables the ability to use advanced queries on the stored information for which query languages such as SPARQL 47 and Cypher 53 have been developed 3 3 1 Inventory An inventory of databases are described in this subsection Each database has advantages and disadvantages that will be used as the input fo
102. o use the standby mode of the monitor which still uses a small amount of energy but the upside is that it will cost nearly half of the first scenario Kawamoto et al 8 Table 5 show that an average monitor uses 1 5W LCD to 5W CRT in standby mode which is automatically applied by the operating system when a computer is put in S3 commonly referred to as standby sleep or suspend to ram S4 commonly referred to as hibernate or suspend to disk or off Kawamoto et al 8 Table 3 also show that a computer and monitor have an average power on time of 6 9 hours per day during office hours including 3 9 hours of idling time Studies also show 8 Table 4 9 Table 3 that a large percentage of monitors during non office hours are not turned off but in standby mode 4096 in Japan and 7596 in US From the observed patterns in 8 a monitor could be in standby mode for 21 hours a day With CRT monitors discontinued for years the LCD monitor is widely used in offices With 21 hours per day in standby mode this results in a total of 0 0315 kWh potential waste each day With energy prices per 12 KoSyary S N t BR j vB xy qpSt OCY 2yXP NCPAC NJdEKT j j RB St OCR t s SySNje Gy dY Ley 20 73 SyaNy At Nj GAB 46 2088 OTS Y 219 kt y ri ES NY 2 ON even resulting in a low return on investment To reduce complexity and due to the low return on investment of the second scenario the decision was made to use the first scenario To measure co
103. often desirable to encrypt the messages that are sent between the components to avoid eavesdropping Perhaps message encryption at application level can be used for the messages that are sent using the message queue e g using public key cryptography where the Kafka producer can use the public key to encrypt the messages whereas the Kafka consumer can decrypt the messages using the private key The current implementation of the Green M ind system allows all messages to be sent from all components As future work it would also be important to implement a more sophisticated access system e g ACL that does not only limit which messages can be sent by each component but also features multiple users with manageable permissions that specify what operations are allowed to provide the system with an additional level of security against unwanted visitors 7 5 Administrator panel The independent configuration and maintenance of the components sensors and actuators is possible Administrators have to manually monitor the state of the components and configuration of these components is done locally through settings files Although maintenance does not require recompiling it is by no means as easy as it could be As a first step it would be of help to have statistics for each component sensor and actuator e g uptime downtime or cpu usage for components to determine their current status As log files have proven useful in helping to maintain the com
104. ogies There are three different configurations for the DEX licenses available 80 Size small up to 1 million objects medium up to 100 million objects large up to 1 billion objects and very large more than 1 billion objects Licenses can be upgraded to a higher size Sessions the number of concurrent sessions between 1 and unlimited High availability provides horizontal scalability that allows DEX to handle larger workloads There are five different licenses of DEX available that have the same amount of features depending on the selected configurations 80 Evaluation license for personal testing purposes The configurable size is limited to small only Research license for academic research purposes All configurations are available but registration is required Development license for business testing purposes All configurations are available but registration is required Commercial license for business production purposes All configurations are available Price 475 EUR per year for the most basic configuration to 11 515 EUR per year for the most complete configuration Support hours are based on the configurable size and professional support is available 8 hours a day 5 days a week 81 Advantages Great documentation available such as guides tutorials and examples 109 Development is widely focused on multiple aspects of the database 85 Efficient usage of available d
105. oom does not leave much room for energy saving There are no fixed devices and users bring their own laptops and a beamer for presentations and meetings Usage is sporadic and unpredictable as can be seen in Figure 5 8 On the individual days the users on Thursday y use more energy in a short period of time but the users on Thursday x used the room for the whole day and consumed less per hour but more in total Taking everything into account the meeting room is an interesting location for monitoring the energy consumption but not for controlling the energy consumption 5 2 Estimated energy saving potential and other research In order to discuss the measurements and the potential savings they entail when using the functionality of the Green Mind system over a longer period of time Table 5 1 has been created to summarize all the findings It should be noted that measurements were taken over a period of six weeks and not over a whole year The current estimated annual values are calculated based upon the average energy consumption of the actual measured devices within and outside office hours during these six weeks whereas the after estimated annual values are estimated only based on the possibilities of the Green Mind system to limit the energy consumption 53 73 Current After Potential annual savings Potential estimated estimated kwh Savings annual kWh annual kWh Light 1 office fixture 136 8 59 85 76 95 56 Light 1 corridor fi
106. other hand the monitored offices belong mostly to the same research group and there are multiple research workstations active 24 hours a day and 7 days a week Depending on the research it might be possible to combine some of these workstations Only eliminating one of these research workstations would already save a considerable amount of energy The decision to turn off a research workstation or at least put in into sleep mode is not something that the system can easily make on its own and requires input from the responsible researchers Most likely the use of research workstation specific rules would be a good compromise For example a research workstation could be turned off at night but during the day including weekends the workstation is expected to be operational The detection of such research devices at least provides building management the possibility to open such discussions with staff members to determine a reasonable policy regarding such devices For a regular workstation the monetary savAy5d 6 AE OS NIdyR Bh Xj tht KMB Tor S 22 f Mid S NI Sy s Akay by ES NY KE 2dtR RR dij NIdyR Bako KXK y2 Sy2dzK CR BNIKS cost of the installed Plugwise device Because saving on the remaining energy consumption if the workstation is turned off alone is not sufficient and to reach the goals stated in the introduction it is necessary to also execute additional algorithms to put computers into standby modes as fast as poss
107. ow down the scope a fixed number of research goals have been set This section describes those goals in Subsection 1 2 2 as well as the plan to reach those goals in Subsection 1 2 3 and the initial expectations about the outcome of the thesis in Subsection 1 2 4 1 2 1 Hypothesis The purpose of this experiment is to design and build an actual working system that is capable of automating energy consumption management in regular offices that are being used by employees Previous research 4 has indicated that energy can be saved by implementing such automated systems but are usually limited to a small scale implementation in a controlled environment The ever growing amount and implementations of sensors and the connectivity of such devices make it possible to implement such a system in a working environment outside of a lab setup By implementing the system in an uncontrolled environment of actively used offices it will presumably be less stable due to users tampering with the sensors out of curiosity or inexperience It also raises the case that no office worker is alike which means the amount of potential energy saving will differ from office to office It will therefore be of use to find out if an automated system will scale up to a larger and more volatile operating environment The system should not lose too much of its effectiveness in order to save energy while justifying the costs both in money and energy consumption of the system itself and
108. ponent rather than the desired form as discussed in Subsection 4 3 2 and the context component which is unable to communicate with other components but can process variable changes and update the context variables accordingly in test situations These components were co developed by two other groups that were unable to finish the components in the time slot of this thesis Although it is currently not possible to test the Green Mind system over a long period of time to determine whether it is able to lower energy consumption completely autonomously the Green M ind system has already been gathering consumption data for weeks and is installed and functioning in the offices All the sensors and actuators can be accessed through the software The controller has been successfully tested with a dummy context component and with that the rest of the system functioned as dictated by the architecture design Automation cannot function due to the incomplete SAGW and context components but the purpose of designing and building a system that is able to actually function in offices has been met By installing the Plugwise devices that function as both sensor and actuator the Green Mind system can reach all the necessary sensors and actuators through the wireless ZigBee network The computer clients also 43 73 communicate through the local TCP network and also require no additional infrastructure to be installed as was intended The computer clients even funct
109. ponents and to investigate problems one area of future work is to combine their information with the statistics to provide easier means of access However only a small subset of all possible behaviors is currently represented in the log files Extending the log files to include more behaviors will provide a wider variety of information for administrators to investigate 61 73 A next step would be to add notifications which requires monitoring and acting upon the reported statistics For example if a component sensor or actuator is down an administrator is able to receive a notification to investigate the cause In case of the components is perhaps also possible to remotely start stop or restart each individual component In the end the panel should be automated to the point that it is not always required to have physical access to the system and that downtime of a component sensor or actuator is kept to a minimum Another area of interest is the more sophisticated access system that was described in the Section 7 4 The nature of such an access system would require a panel to manage users with manageable permissions as well As such combining the access system panel with the administrator panel may help the user by providing a single point of entry to all administrative tasks By using a web interface to access the administrator panel it could provide access to nearly all users In addition a mobile client for platforms such as Android iOS
110. ported For an implementation with the scale such as the Green M ind system it might even turn out that home built sensors are able to provide a better and cheaper solution The main concern is that installing all kinds of sensors is expensive and potentially unnecessary If done correctly the installation and choosing of the sensors is a whole separate research project All in all the Green M ind system is capable of reaching the goals set out in the introduction Both lighting and devices can be monitored and controlled The Green M ind system is a framework that allows the University of Groningen to expand its research into sustainability in new direction such as user awareness and provides the researchers with real and up to date data The loose design of the system components means that it can be enhanced over time exceeding the goals of this initial design and thus provides more possibilities than the existing built in BM S 44 73 Chapter 5 Theoretical results Chapter 5 presents the collected energy consumption measurements from the Green Mind system in the office environment that was discussed in Section 4 1 First the measurements are presented as is in Section 5 1 The measurements presented in Section 5 1 are without the system s ability to automatically influence the energy consumption Using the gathered energy consumption measurements the estimated energy saving potential of the system is discussed in Section 5 3 based on the
111. ps community jboss org wiki HornetQGeneralFAQs REDHAT Red Hat Subscription model visited 29 07 2013 http www redhat com about subscription ab1 RABBITM Q RabbitM Q Highly Available Queues visited 29 07 2013 http www rabbitmg com ha html APACHE Apache Kafka visited 29 07 2013 http kafka apache org 08 quickstart html JBOSS Chapter 38 Clusters visited 29 07 2013 http docs jboss org hornetq 2 2 5 Final user manual en html clusters html 69 73 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 APACHE Apache Kafka visited 29 07 2013 http kafka apache org design html JBOSS HornetQ User M anual visited 29 07 2013 http docs jboss org hornetg 2 3 0 Final docs user manual html single index html APACHE Index Apache Kafka Apache Software Foundation visited 29 07 2013 https cwiki apache org confluence display KAFKA Index APACHE Apache Kafka visited 29 07 2013 http kafka apache org index html APACHE Apache Kafka visited 29 07 2013 http kafka apache org 07 quickstart html JBOSS HornetQ QuickStart Guide visited 29 07 2013 http docs jboss org hornetg 2 3 0 Final docs quickstart quide html single index html RABBITM Q RabbitM Q Getting started with RabbitM Q visited 29 07 2013 http
112. r the eventual approach The most described graph database choices are 76 91 92 93 Neo4j 117 Titan 118 HyperGraphDB 119 AllegroGraph 120 DEX 121 InfiniteGraph 122 OrientDB 123 m u I I I The following graph databases were dropped for consideration after an initial review Titan due to the large number of issue reports 94 third party licensing problems 113 and an unpredictable fluctuating performance 114 HyperGraphDB due to an inadequate community 49 no replication mechanisms 96 and the time between subsequent releases being nearly 2 years 97 98 AllegroGraph due to only supporting 5 million nodes in the free version 99 an inadequate community 101 support that is accessible only through email e g no forums 100 and a focus on describing and interchanging metadata on the web i e Resource Description Framework RDF information rather than a general purpose database 102 InfiniteGraph due to the pay as you go pricing starting at 4 000 EUR the price depends on the number of supported nodes and edges 103 an inadequate community 104 and no replication mechanisms 105 The three remaining graph databases are subject to a closer review Neo4j is described in Subsection 3 3 2 OrientDB in Subsection 3 3 3 and DEX in Subsection 3 3 4 23 73 3 3 2 Neo4j Neo4j is a robust and fully ACID transactional Java persistence engine that stores data structured
113. raph Database visited 30 04 2013 http www neotechnology com price list NEO4J Neo4j the Graph Database Language Drivers visited 30 04 2013 http www neo4j org develop drivers NEO4 J Neo4j the Graph Database Ask Questions and Share Answers visited 30 04 2013 http www neo4j org participate q and azstack overflow NEOTECHNOLOGY Customers Neo Technology Neo4j Graph Database visited 30 04 2013 http www neotechnology com customers NEO4J Neo4j the Graph Database Learn Cypher the Neo4j query language visited 30 04 2013 http www neo4j org learn cypher NEO4 Appendix B Questions amp Answers visited 30 04 2013 http docs neo4j org chunked stable questions html NEO4J Neo4j Blog 2013 What s Coming Next in Neo4j visited 30 04 2013 http blog neo4j org 2013 01 2013 whats coming next in neo4j html NEO4J 25 2 Performance Guide visited 30 04 2013 http docs neo4j org chunked milestone performance quide html NEO4 J Chapter 22 High Availability visited 30 04 2013 http docs neo4j org chunked stable ha html WEBBER Scaling Neo4j with Cache Sharding and Neo4j HA World Wide Webber visited 30 04 2013 http jim webber name 2011 02 scaling neo4j with cache sharding and neo4j ha NEO4J Neo4j the Graph Database Other Neo4j Versions visited 30 04 2013 http www neo4j org download other versions NEO4J Neo4j the Graph Database Learn visited 30 04 2013 http www neo4j o
114. red to the second scenario The main shortcoming of the second scenario is that acting upon detecting when the computer is not being used cannot happen instantly For example when only input devices are used for computer usage detection it is unlikely that the user is constantly using the input devices of the computer A delay e g 10 minutes can partially solve this although a delay will never achieve the same accurate results as using the first scenario However as the first scenario already costs MdaKt8 rij B 26 AK2d iK Mi dat EY LHN TS HS yia NIE yR 050 da Ais 2dtR AME ySSR RSE Sep the user is only grabbing a piece of paper from a drawer the decision was made to use the second scenario 3 2 3 Lighting Lights are used throughout the whole building not only the office lights can be influenced but also the lights in the hallways For the lights in the building there are three variables to be measured to determine context and to influence the energy consumption Movement to determine context and influence the energy consumption The movement can be used to determine whether lights should be switched on or off Energy consumption where the measurement for determining context can either be done over a single light or multiple lights at the same time A baseline measurement of the energy consumption compared to later measurements can measure how effective the solution is Light intensity where the value can be used to determine w
115. rg learn ODINO Graph databases OrientDB to the rescue Alessandro Nadalin visited 01 05 2013 http odino org graph databases orientdb to the rescue ORIENTECHNOLOGIES Orient Technologies the company behind OrientDB visited 01 05 2013 http www orientechnologies com ORIENTECHNOLOGIES Orient Technologies the company behind OrientDB visited 01 05 2013 http orientechnologies com enterprise htm ORIENTECHNOLOGIES Orient Technologies the company behind OrientDB visited 01 05 2013 http orientechnologies com support htm 65 73 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 TINKERPOP TinkerPop visited 01 05 2013 http www tinkerpop com GITHUB Programming Language Bindings nuvolabase orientdb Wiki GitHub visited 01 05 2013 https github com nuvolabase orientdb wiki Programming Language Bindings GITHUB Production Deployments nuvolabase orientdb Wiki GitHub visited 01 05 2013 https github com nuvolabase orientdb wiki Production Deployments GITHUB Tutorial Document and graph model nuvolabase orientdb Wiki GitHub visited 01 05 2013 https github com nuvolabase orientdb wiki Tutorial Document and graph model GITHUB Replication nuvolabase orientdb Wiki GitHub visited 01 05 2013 https git
116. round 167 watt when all processing power is being used On average the consumption is just above 100 watt The slight dip around 13 00 is most likely caused by the staff member going out for lunch and manually turning off the monitor and or computer for the duration of the lunch Note that the workstation still consumes some energy outside office hours The week graph in Figure 5 2 shows the consumption of workstations throughout the week Workstation 1 WS1 is similar to the workstation from the top graph However workstation 2 WS2 is used for research purposes and thus never shuts down not even in the weekend WS2 poses the problem that workstations for research are usually stocked with above average hardware to be able to process high workloads that usually requires more energy to do so In the monitored office environment there are multiple machines that are 47 73 powered on continuously all belonging to the same research group The Green Mind system could put these devices into standby mode when they are not being used such as weekend and nights but as those systems are used for research purposes this might not always be appreciated by the staff Instead it could be enforced as a university policy that only designated machines can be used for long term tasks instead of allowing every researcher to run his own local server or monitoring system Note that the energy consumption of WS2 stabilizes in the weekend at the lowest energy consumption
117. s A denotes an exceptional property a denotes a positive property whereas a denotes a negative property for a specific message queue Regarding the scalability and performance guarantees for the Green Mind system there are no specifics about the actual requirement for these guarantees due to several reasons The main reasons are that the amount of messages highly depend on the number of sensors and actuators currently undecided how often the measured quantities change from the sensors and the deployed scale of the system 30 73 Regardless the scalability and the performance are an important aspect of the system especially when it is likely that one message queue instance will be used per building or even for a whole university gna 1 gna Scalability and Can send to master Can send to each node Can send to each node Availability Amount of configuration Amount of configuration Requires shared storage Amount of configuration Performance 2 1 3 Second best performance Best performance Worst performance compared to Kafka and compared to RabbitMQ and compared to RabbitMQ and HornetQ HornetQ Kafka Documentatio 1 ze a n Best documentation Great documentation Worst documentation available available available Community 1 3 25 Largest and most active Smallest but active Substantial and active community community community 1 1 Qn Used by large sized Used
118. s on not operated by movement like the office lights Figure 5 6 Energy consumption summary of lights in the corridor The two hallways of the monitored office environment are so large that the two hallways have been divided into three corridors with a Plugwise device monitoring the corresponding corridor The hallway lighting has the same hardware as the office lighting but the hallway lights are not controlled through an PIR sensor Lights are continuously lighting the hallways until they are turned off on a building level As they are similar to the office lighting they also have the same flaw on sunny days they remain turned on Figure 5 6 shows the energy consumption of the different corridors The consumption pattern of the corridors is identical but the amount of fixtures in corridor 2 and corridor 3 is slightly lower than the amount grouped into corridor 1 Both hallways consume around 1900 watt combined when turned on This means that on a sunny day each hour 1 9 kWh in the monitored office environment on the fifth floor alone could potentially be saved in energy consumption simply by turning the hallway lighting off Taking into account that the hallway lights are continuously turned on when it is sunny outside there is a total energy saving potential of 1846 8 kWh 36 per year with an average of 4 5 hours of sunlight per day and 228 workdays per year In some cases it might be desired to have more detailed control over the lighting so t
119. sensors and actuators to monitor the energy consumption of devices and to turn them on or off A computer client has been developed that acts as an activity sensor and provides the system with the ability to safely shut down computers or to put them into a standby mode The installed system consists out of multiple loosely coupled software components that each fulfill their own dedicated tasks Kafka has been chosen as the message queue by which the components communicate with each other through JSON messages A ZooKeeper instance is used to monitor the state of components and provide automatic discoverability for all the components including the Kafka message queue Based on the actual measured energy consumption data and the functionality of the created Green Mind system a simulated result was formed for the energy saving potential of the design The biggest energy saving potential is the office and hallway lighting which even on a bright sunny day consumes multiple kilowatt hours per day Computers use around one kilowatt hour per week per computer outside office hours even when turned off Through a smart set of rules the Green Mind system is capable of saving that energy by automating the control of the devices Acknowledgments We would like to express our sincere gratitude to our supervisor Alexander Lazovik for his continuous support enthusiasm and knowledge Without his guidance and persistent help this thesis would not have been possibl
120. stems are widely used in buildings to observe and grasp the status of energy consumption in operation Besides monitoring BMS systems are able to control the internal environment of the building and can if configured properly reduce the energy consumption However these BM S systems typically do not have fine grained access to each room or even each floor and encompass a larger area For example the BM S used at the Bernoulliborg building only has access to a whole wing and cannot monitor individual rooms Studies 17 18 19 20 21 have been presented by the scientific community to incorporate more intelligence in the BMS systems to optimize the building s energy usage Two studies 17 21 show a similar effort to reduce the energy consumption by equipping the building with multiple sensors and provide that information to the BM S system Rules are created based on the context information provided by the sensors to achieve an optimal energy efficiency and comfort condition of the building Although both studies seem to be one of the most well described and complete approaches that are similar to the work in this thesis both require a BMS The approach in this thesis does not depend on a BMS system and can therefore be used in any building In addition the study presented by Han et al 21 does not provide any information about potential energy savings that can be achieved through the system The study by Doukas et al 17 only provides information
121. t can request the energy consumption and also change the state of the Plugwise devices i e on off 3 1 2 Service layer Components in the service layer use the functionality of the physical layer to provide added value for the higher layers The service components provide functionality that exceeds the get and set level of the physical layer and simplifies sensor and actuator access even further Some components create new output themselves e g when processing and converting the consumption data from the Plugwise devices while others just process high level commands and translate those to commands that can be used by the SAGW For example the context component creates new variables e g room occupancy by combining information received from multiple sources such as the sleep management server SM S and the SAGW The SMS enables through the SAGW PC client the orchestrator to manage the state of the connected computers and provides information about activity on the computers to the context component The orchestrator is also able to change the state of the Plugwise devices through the SAGW Plugwise interface The consumption component uses the SAGW Plugwise interface also namely to measure the energy consumption of the devices that are connected to the Plugwise devices 3 1 3 Persistence layer The persistence layer acts as a system wide long term storage The data is stored by the components in such a way that it can easily be accessed
122. t of the devices that are common goods These devices are used by multiple people and mostly remain active throughout the day even without people being present e g hallways are always lit 5 1 1 Offices Two types of devices are measured in offices There are the lights which are grouped together to a single Plugwise device in case of multiple fixtures and the workstations which have one Plugwise device each Due to modern monitors already having a low energy consumption in standby mode as discussed in Subsection 3 2 2 it was decided to group the computer monitor and any auxiliary devices in a power strip that is connected to the workstation Plugwise device The Green Mind computer client can put the computer into Sleep or hibernate mode The Plugwise device only needs to be used as an actuator if it is desirable to eliminate any remaining standby power consumption of the workstation completely 45 73 si M easured between 75 225 watt 75 watt per fixture r 9 z Lights 2 200 Light 150 100 5 50 0 E ws T T T T T T 01234 5 6 7 8 9 101112131415 16 17 18 19 20 21 22 23 Hour of day Common consumption pattern of two combined light fixtures in the course of one day Lights 1 2 3 4 5 6 7 Day of week Mon Sun Common weekly consumption pattern of two combined light fixtures in an office Notes Maximum of three light fixtures per office
123. ten Damen Plugwise Unleashed 0 1 A document explaining the protocol used by Plugwise products 2010 DIRKENGELS PlugW eb Creating the driver Dirk Engels visited 07 08 2013 http wordpress dirkengels com 2011 04 24 plugw eb creating the driver Georgievski D Degeler G A Pagani T A Nguyen A Lazovik and M Aiello Optimizing Offices for the Smart Grid University of Groningen JBI 2011 12 01 2011 ANTLR ANTLR visited 15 08 2013 http antlr org SINGULARSYS Jep Java Math Expression Parser Singular Systems visited 15 08 2013 http www singularsys com ep APACHE OGNL Apache Commons OGNL Object Graph Navigation Library visited 15 08 2013 http commons apache org proper commons ognl CODEHAUS M VEL Performance of MVEL 2 0 visited 15 08 2013 http mvel codehaus org Performance of M VEL 2 0 APACHE Commons JEXL Overview visited 15 08 2013 http commons apache org proper commons jexl ANTLR Getting Started with ANTLR v4 ANTLR 4 ANTLR Project visited 15 08 2013 http www antlr org wiki display ANTLR4 Getting Started with ANTLR v4 SINGULARSYS Trial Download Singular Systems visited 15 08 2013 http www singularsys com jep download trial php SOURCEFORGE Java Unified Expression Language visited 15 08 2013 http juel sourceforge net GITHUB Commits beckchr juel GitHub visited 15 08 2013 https github com beckchr juel commits master APACHE Apache S
124. tet ro eoe 15 S 1 DESION OVELVIEW N oot dor ton eor o m eot dot th o Ted Mer 15 31 T A c t err eret aen ep in tete ei ef o ed t bien e dede 16 3 1 2 Service layer iii an eee emi pa e sid t aei te rer eden en ada ee p dun 16 3 1 3 Persistence layer i need Gee tpa e niit aile e Pe A 16 3 14 Logic ayer oti A t EHE er EOD del Le ear ated ER ek HA d 17 3 2 Sensarsand actuators nee onse dare re ne DUE E Peek eR La ee 17 3 2 T INVENLONY EN Mla aR ete E Ain ERR RR Ee i asa 17 3 2 2 COMPUTE Teresi ence iet eer ite c Le d n A A teins AR ctt e Led ee date ehe 20 3 2 3 Te yii eR P AE eher re er re ae TREE ENTE ne eier 21 O en cd Haee ir mtt p Eco pde UR 22 3 3 Databases is AA rede de ee e HU n UTERE C DOR Rr UU e ERA e 23 S PS INVENIA 23 3 32 N60 P 24 3 3 3 OOO ii A E E 24 3 34 DEX IENEI EER EAE PAT E T AENT 25 3 3 9 OI iii di ap ob o oco ie d b OR Cr e T D n E a a EB 26 3 3 3 CO irradia eio e ro d d Y ea ep Hp pO I rc ag EB E 27 344 Message queue ax n e tn e kan A a a hed E ean dln 28 SN eer terere ee enge 28 3 4 2 RabDItM Q iarann edi e UR RU I ELTERN on ATAA Du EOI 29 3 4 3 Kafka rere rere pu Dee cet ut EU E dte qo ur prd quet eee pes 29 3 44 HottietQ atitem iter teet i ete Pad ub eta nit a a d e ad uen 29 AO iran e tuer pn e PR Rr EROR TR ORG ORTA LEVEL en ESTHER rk Te TORRE adobe 30 3 4 6 COnelUSIOn usa pU PE d ERE FR REO E UTR Rd ASA rU
125. than the message queue can process it will eventually come to a stop Documentation The documentation of the message queue can address topics like tutorials how to setup the message queue and tutorials A message queue is not considered without well written and complete documentation as there are too many uncertainties and risks involved Community An important aspect of a message queue is the community evolving around it Without a community it is often hard for users to ask questions and find tutorials about the product A community is also important to ensure that development continues with input from the users e g suggest new features or submit bug reports Usage The usage or adoption rate is an important aspect of a message queue It affects aspects such as development e g continue with input from the users or the possibility to add patches or functionality in open source software stability e g it is unlikely a company is going to use unstable software in production and support e g more users available to ask and answer questions For each of the six guarantees there is one message queue to provide the 1 best solution with the other two message queues providing the 2 or 3 best solution The guarantees are compared based on the closer review from subsections 3 4 2 3 4 3 and 3 4 4 and are shown in Table 3 2 The scalability and availability guarantees are grouped together in Table 3 2 because of their related propertie
126. tion pattern of the microwave oven in the course of a day Microwave 1000 E 500 0 T T T NEN 1 2 3 4 5 6 7 Day of week Mon Sun Common weekly consumption pattern of the microwave Notes Uses around 2 6 watt when idle Figure 5 3 Energy consumption summary of the kitchen microwave oven 48 73 The microwave is used throughout the day and energy consumption peaks during lunch Figure 5 3 shows the energy consumption of the microwave in detail The energy consumption of the microwave can not be called excessive around 2 6 watt is consumed when the microwave is idle By turning the device completely off outside office hours around 274 watt 2 6 watt 105 5 hours can be saved per week which still adds up to around 696 saving of its weekly energy consumption M easured between 38 192 watt E Coffee machine 200 Coffee machine e pim 50 Q 4 r r 0123456 7 8 9 101112131415 16 17 18 19 20 21 22 23 Hour of day Common consumption pattern of the coffee machine in the course of one day Coffee machine 2000 1500 A EE 1000 500 0 T T 1 2 3 4 5 6 7 Day of week Mon Sun Common weekly consumption pattern of the coffee machine Notes Consumes 56 watt on average when idle Turning it completely off mig
127. trieve the consumption measurement data it is not part of the SAGW Retrieved data needs to be correctly stored in the database and old or missing history log entries also need to be retrieved This added data management is not part of the SAGW and thus are the energy consumption measurements handled by this separate component Plugwise stores data in the form of an amount of pulses which can be converted to the corresponding watts being consumed by the connected device s Figure 4 3 shows all the data in an hour entry that is returned by the SAGW after requesting the energy consumption of a Plugwise device M ost of the values are derived from the pulses and are stored to simplify processing for higher level components Some sources 185 186 indicate that the pulses should be corrected for both the current energy consumption and the historic energy consumption whereas another source 187 indicates that the corrections are only used for the historic energy consumption However the software developed by the manufacturer i e Plugwise Source is only using the uncorrected values As there is no definitive answer it was therefore opted to return both values to minimize problems with conversions in the future All returned values are stored in the database to make sure that no data is lost address 0DOGOEAC time 2013 06 03T15 00 00 000402 00 pulses 974116 watt 577 0219 kwh 0 57702184 correctedpulses 558222 6 corr
128. ty on energy and sustainability and also the implementation of the winning plan of the Green Mind Award Through this thesis a framework is built that presents a multitude of new research possibilities for bachelor and master thesis projects such as research into energy awareness amongst employees as well as allowing the university to design and develop new homegrown ideas a desire expressed in the original call for submissions for the Green Mind Awards 5 Accessibility to a live testing environment of actively used offices is something that is rarely done before and gives the University of Groningen an edge over other universities which they can use as a selling point Other research focuses on measuring and influencing devices in a controlled environment and point out that there is energy to be saved through smart automated management This project aims to determine if such solutions are still manageable and efficient in a larger and more volatile environment Another point of focus is to create a system that is setup in such a manner that it is applicable to multiple buildings Not just another prototype but a system that is usable in production in the future Although the topic of sustainability is hot at the moment research has yet to bring forth a fully functioning solution that can operate in a working environment This thesis introduces a new system design and contains the consumption data collected from the offices located at the University
129. uch a way that it can easily be accessed in the future The storage can also be used by components to store their state in so that a component can always continue from the previous state if necessary e g in the case of a hardware failure The separation of the storage into the persistence layer makes it easier to manage the storage for all the components combined instead of each component managing their own storage system Another advantage of using one storage is the potential to increase availability and stability of the system by adding more servers as it is possible that individual components will not expand their storage system unless absolutely needed e g performance of an individual server becomes too low Figure 3 1 shows the persistence layer above the physical and service layer Typically the persistence layer is one of the lowest layers because the only purpose is the storage and retrieval of data Figure 3 1 differs from the typical order merely for the sake of keeping the design of the figure easy to understand while still showing all the connections to the storage Based on importance and functionality the persistence layer could be placed at the same level as the physical layer 4 2 4 Logic layer The components belonging to the logic level are the high level components These components use all the available data and functionality from the lower layers in order to make calculated decisions about the state of devices This is also
130. uropean low energy office buildings A literature review Energy and Buildings Volume 43 Issue 10 October 2011 Pages 2572 2582 H B low H be Daylight in glazed office buildings a comparative study of daylight availability luminance and illuminance distribution for an office room with three different glass areas Report EBD RG 08 17 Lund University Department of Architecture and Built Environment Division of Energy and Building Design Lund Sweden 2008 N Borg Guidelines for integrating sustainable summer comfort into public procurement schemes for office equipment and lighting Keep cool program Deliverable 3 2 Swedish Energy Agency October 2009 D Loe Energy efficiency in lighting jan overview Action Energy GIRO92 Society of Light and Lighting London 2003 M Santamouris A Argiriou E Dascalaki C Balaras A Gaglia Energy characteristics and savings potential in office buildings Solar Energy Volume 52 Issue 1 January 1994 Pages 59 66 A Boyano P Hernandez O Wolf Energy demands and potential savings in European office buildings Case studies based on EnergyPlus simulations Energy and Buildings Volume 65 October 2013 Pages 19 28 Yuvraj Agarwal Bharathan Balaji Rajesh Gupta Jacob Lyles M ichael Wei and Thomas Weng 2010 Occupancy driven energy management for smart building automation In Proceedings of the 2nd ACM Workshop on Embedded Sensing Systems for Energy Efficiency in Building BuildSys 10
131. vices Plugwise also notes that the one and eight seconds buffers are not as reliable as the hourly intervals and the hourly intervals should always be preferred for precise measurements The consumption component uses the history buffer instead of the one or eight second buffers to limit the network traffic and processing load Upon request the log address of the log that contains the desired hour e g the previous hour is passed along Each log address points to four entries of an hour e g when requesting the consumption for 14 00 a response could contain 11 00 12 00 13 00 and 14 00 depending on the log address used for the request The consumption component requests the historic data once every hour from the SAGW and continues retrieving entries until the last history entry that is stored in the database is reached By requesting all history logs until the last log that was stored in the database it is also no longer an issue when the consumption component fails the Plugwise devices will continue to monitor the energy consumption and when the consumption component restarts it will simply gather all the missing data The same also applies when the network of the Plugwise devices is temporarily unreachable When the network connection is 37 73 reestablished the consumption component will retrieve everything until the last stored history log in the database Even though the consumption component works closely together with the SAGW to re
132. visited 22 07 2013 https github com orientechnologies orientdb wiki SPARSITY TECHNOLOGIES Sparsity technologies DEX high performance graph database visited 22 07 2013 http sparsity technologies com dex tutorials GOOGLE Downloads orient NoSQL document database light portable and fast Supports ACID Tx Indexes asynch queries SQL layer clustering etc Google Project Hosting visited 22 07 2013 http code google com p orient downloads list GITHUB Releases orientechnologies orientdb GitHub visited 22 07 2013 https github com orientechnologies orientdb releases 67173 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 TWITTER SparsityTechnologies SparsityTech on Twitter visited 22 07 2013 https twitter com sparsitytech GOOGLE Licenses Google Groups visited 22 07 2013 https groups google com forum searchin aureliusgraphs license aureliusgraphs 6fG6BaGDFEO GOOGLE performance in Aurelius Google Groups Search visited 22 07 2013 https groups google com forum searchin aureliusgraphs performance TOYOTARO SUZUM URA AND MIYURU DAYARATHNA 2012 XGDBench A benchmarking platform for graph stores in exascale clouds In Proceedings of the 2012 IEEE 4th International Conference on Cloud Computing Technology an
133. w rug nl research energy Georgievski Ilce and Degeler Viktoriya and Pagani Giuliano Andrea and Nguyen Tuan Anh and Lazovik Alexander and Aiello M arco Optimizing Energy Costs for Offices Connected to the Smart Grid In IEEE Trans Smart Grid 3 4 2273 2285 Year 2012 RUG University of Groningen awards Green M ind Award 2012 for the most sustainable idea University of Groningen visited 17 07 2013 http www rug nl news and events news news2012 object974926682 lang en D Kolokotsa K Niachou V Geros K Kalaitzakis G S Stavrakakis M Santamouris Implementation of an integrated indoor environment and energy management system Energy and Buildings Volume 37 Issue 1 January 2005 Pages 93 99 A Guillemin N Morel Experimental results of a self adaptive integrated control system in buildings a pilot study Solar Energy Volume 72 Issue 5 M ay 2002 Pages 397 403 Kaoru Kawamoto Yoshiyuki Shimoda M inoru M izuno Energy saving potential of office equipment power management Energy and Buildings Volume 36 Issue 9 September 2004 Pages 915 923 ISSN 0378 7788 Carrie A Webber Judy A Roberson Marla C M cWhinney Richard E Brown Margaret J Pinckard John F Busch After hours power status of office equipment in the USA Energy Volume 31 Issue 14 November 2006 Pages 2823 2838 ISSN 0360 5442 Marie Claude Dubois Ake Blomsterberg Energy saving potential and strategies for electric lighting in future North E
134. wered on but not used and the active time i e the computer is powered on and used The idle and active time are both important to determine context and to influence the energy consumption The following sensors can be applied for the energy consumption and computer usage Energy consumption sensor To monitor the energy consumption of the computer monitors or both combined Ultrasonic range sensor To determine if someone is sitting in front of the desk Computer software sensor To detect computer usage and to influence the power state sleep hibernate or off which hasthe potential to affect the energy consumption Seat pressure sensor To determine if someone is setting on their seat To measure the energy consumption of the computer there are two possible scenarios 1 1x energy consumption sensor and 1x power strip The power strip will be used to connect both computer and monitor s to The energy consumption sensor will measure the total energy consumption of all devices connected to the power strip 2 2xenergy consumption sensor and 1x optional power strip The power strip will only used when a computer has multiple monitors The energy consumption sensor will measure the energy consumption of the computer and monitors separately The advantage of using the second scenario is that it will be possible to turn off a monitor completely using the energy consumption sensor With the first scenario it will only be possible t
135. xture 213 75 136 8 76 95 Due to diversity only the standard university issued workstation is taken into account Research workstations consume more power and are active 24 hours a day and 7 days a week partial down time will at least equal or exceed the savings of a standard workstation Table 5 1 Summary of the potential energy saving found through the measurements 5 2 1 Lighting As mentioned in subsection 5 1 2 the hardware of the corridor and office lighting is the same This means that the potential savings are also the same when disabling the lights when the sun is out For an office fixture this means that around 5696 of the annual energy consumption can potentially be saved For the corridor lighting this is only 3696 because the building remains open for longer than the regular office hours and unlike the office lighting the corridor fixtures are not controlled by a PIR sensor Corridor lights are thus turned on longer annually while the hours of sunlight are the same for both resulting in less potential savings The savings of 3696 and 5696 are substantial because they could be reached even when the installed fixtures already contain energy efficient lighting and offices are even only enabled by motion sensors According to Boyano et al 15 lighting is the highest contributor to energy consumption in medium climate zones such as the Netherlands at 5096 Through lighting control such as taking natural light into account they
136. y coupled component the energy saving potential of the Green Mind system is only expected to increase over time when old components are replaced with newer more sophisticated components By providing a framework that can be built upon more time can be put in working out the details of individual components and less time worrying about the system architecture itself 58 73 The estimates show that the following starting hypothesis is certainly true Installing a building automation system that controls multiple occupied offices has the potential to save energy A lot of interesting research is still to be done in the field of building automation and by having a platform such as the Green M ind system the University of Groningen is able to advance into the next stages of saving energy through automation 59 73 Chapter 7 Future work Although the results presented have shown energy saving potential by using component based automation it could be further developed in a number of ways 7 1 Extending and improving the sensors and actuators In principle the proposed sensors and actuators are able to provide the energy saving potential for the system However limitations on the proposed sensors and actuators have been discussed in Chapter 3 Perhaps by incorporating other types of sensors and actuators such as RFID NFC or WiFi tags to detect occupancy and user location it is possible to obtain more accurate results and enhance the

Office energy saving potential through component based automation

Contents

Download Pdf Manuals

Related Search

Related Contents