Kilcher et al adv_tu..
Contents
1. Format the time axis tkr dt MinuteLocator interval 5 frmt dt DateFormatter 3H M ax xaxis set_major_locator tkr ax xXaxis set_minor_1 ax xaxis set_major_formatter frmt ax set_yl Lim Label the ax set_yl abel Oy 31 axes CSu Amat hremi m s 97 ax set_x abel Cmime dune 27 2012 datetime 2012 O Si locator dt MinuteLocator interval 1 size large ax set_title Data Cropping and cleaning 31 K TOONS shrink 0 0 ax set_xlim dt date2num dt datetime datetime 2012 6 12 12 dt dateznum at datetime datetime 2012 6 12 12 380 Save the figure fig usavetig fig crop data pdt end cropping figure Hitt Perform motion correction including rotation into earth frame dat props body2head_vec body2head_vec dat props body2head_rotmat body2head_rotmat mc adv motion CorrectMotion accel_filter mc dat a P Ot dat mp lime datur bo Then rotate it into a principal axes frame adv rotate earth2principal dat adv rotate earth2principal dat_cln Define an averaging object and create an averaged data set binner adv TurbBinner n_bin 19200 fs dat fs n_fft 4096 dat_bin binner dat dat cin bin banner dat cin At any point you can save the data dat_bin save adv_data_rotated2principal h5 And reload the data dat bin Copy2. URL http files microstrain com 3DM GX3 15 25 MIP Data Communications Protocol pdf Nortek August 2005 Vector Current Meter User Manual Vangkroken 2 NO 1351 RUD Norway h ed Polagye B Thomson J 2013 Tidal energy resource characterization methodology and field study in Admiralty Inlet Puget Sound WA USA Proceedings of the Institution of Mechanical Engineers Part A Journal of Power and Energy 227 3 pp 352 367 Priestley M 1981 Spectral Analysis and time series Academic Press Rgstad J September 2011 System Integrator Manual Nortek AS Vangkroken 2 NO 1351 RUD Norway Stacey M T Monismith S G Burau J R 1999 Observations of Turbulence in a Partially Stratified Estuary Journal of Physical Oceanography 29 pp 1950 1970 Thomson J Kilcher L Richmond M Talbert J deKlerk A Polagye B Guerra M Cienfuegos R 2013 Tidal turbulence spectra from a compliant mooring Ist Marine Energy Technology Symposium Washington DC 21 Thomson J Polagye B Durgesh V Richmond M 2012 Measurements of turbulence at two tidal energy sites in Puget Sound WA Journal of Oceanic Engineering 37 3 pp 363 374 Wahl T L 2003 Discussion of Despiking Acoustic Doppler Velocimeter Data by Derek G Goring and Vladimir I Nokora Journal of Hydraulic Engineering 129 pp 484 487 22 A Coordinate systems Tracking coordinate systems referen
3. adv load adv_data_rotated2principal h5 HEHH Figure to look at spectra fig2 plt figure 2 figsize 6 6 palo is clan ax fig2 add_axes 14 14 8 74 axX loglog dat_bin freq dat_bin Suu_hz mean 0 b label motion corrected ax loglog dat_cin_bin freq dat_cln_bin Suu_hz mean 0 te abel no Motion Correction Add some annotations ax axhline 1 7e 4 Color k zorder 21 ax text 2e 3 1 7e 4 Doppler noise level va bottom ha left ax text 1 2e 2 Motion nCorrection axvannotate 2 66e1 Se 3 O 2e 2 arrowprops arrowstyle fancy connectionstyle arc3 rad 0 2 RACE co Moro 32 Fedgecolom 207617 ber ha center ax annotate a mee ve 3 O Be 20 arrowprops arrowstyle fancy connectionstyle arc3 rad 0 2 are co MOG HOR S Zredgecolorm s 1067 e ha center Finalize the figure SS 20N ax set_ylim le 4 1 ax set xillabel tErequencey Mhz ax set_ylabel mathrm m 2s 2 hz size large ma ax set xlim ma f tmp np logspace 3 1 ax plor f tmp le S ss o es lO Y E le gt ax set_title Velocity Spectra ax legend ax axhspan le 4 3e 4 facecolor w alpha 0 8 zorder 10 edgecolor none ax axvspan 1 16 0 2 facecolor 0 8 zorder 10 edgecolor none axX text 4 4e 4 Doppler noise va bottom
4. 04 00 08 00 12 00 Time June 16 2014 Figure 6 A time series of turbulence statistics measured from a TTM at Admiralty Head a velocity b turbulence intensity c turbulent kinetic energy and its components d Reynold s stresses Shaded re gions indicate ebb red and flood blue periods where U gt 0 7 Turbulence intensity is only plotted during these periods because it is meaningless for small values of U The mean over the data record is 10 15 35 1 5 U m s Figure 7 A histogram of the mean horizontal velocity magnitude 0 Fixed ADV Moored ADV N Persistent B motion contamination A Inertial Subrange motion correction m s bz Inertial Subrange Doppler Noise f bz Figure 8 A comparison of the shape of spectra at two different sites from ADVs on a rigid tripod A and a TTM B The spectra for each velocity component u v w are in blue green and red respectively The shaded region indicates the inertial subrange in which the spectra decay like f 5 3 and all components have nearly the same amplitude The dashed line indicates a f gt gt slope The difference in amplitude of the spectra be tween A and B is expected because the turbulence measurements were made at different sites In each panel the doppler noise level arrow points at doppler noise that exceeds the high frequency turbulence levels 16 v spectra w Spectra re OT Tr
5. ET m s hz f bz f nz f bz Figure 9 Spectra of turbulence highlighting motion correction A Streamwise velocity B cross stream velocity C vertical velocity Black lines show the uncorrected spectra red show spectra of head motion and blue shows the spectra after motion correction Green shading highlights loca tions where motion correction reduced the spectral amplitude The uy and v spectra have sharp peaks at 0 1Hz and lesser peaks at higher frequencies that match the spectrum of ii indicating motion contamination Motion correction removed the vast majority of the contamination from these spec tra but contamination from the large peak in the v spectra at 0 1Hz persists The w spectra is essen tially uncontaminated by the mooring motion The inertial sub range is shaded as it was in Figure 8 There are two primary sources of error in moored ADV spectra 1 Doppler noise and 2 imperfect motion correc tion Doppler noise has been studied at length and is easy to identify and account for Doppler noise is a low energy white noise that results from uncertainty in the doppler shift recorded by the ADV In measurements of oceanic turbulence it is generally observed at high frequencies where the amplitude of the turbulent motions drops below the doppler noise level Figure 8 Estimates of S v from a TTM show a peak near 0 1Hz that deviates from the f 23 spectral slope Figure 8B Clos
6. a scalar quantity that includes all turbulence components it has been studied at length by turbulence scientists and it has a well defined budget equation that is the basis of turbulence theory For some purposes it may be useful to investigate each component of F individually Lastly it has sometimes been suggested that a turbulence intensity based on tke that is Ize VErke would be more meaningful to engineering applications but this approach has not gained wide acceptance 3 4 3 Reynold s stresses Reynold s stresses are correlations between velocity components and are fundamentally important to turbulent flow fields Unlike Exe Reynold s stresses appear in the mean flow equation explicitly as terms that transport move momentum from high velocity to low velocity regions Because of how they appear in the mean flow equation Reynold s stresses are typically treated as three distinct components w v ww and ww Several recent studies have found evidence that they are correlated with increased wind turbine fatigue loads e g Kelley et al 2002 2005 which has begun to elevate their importance in the wind energy field 3 4 4 Turbulence auto spectra Turbulence velocity auto spectra hereafter simply spectra are estimates of the distribution of turbulent energy as a function of frequency That is a spectrum quantifies the amount of energy in the velocity at a range of time scales Furthermore since ti
7. adv data dat_raw adv read_nortek fname Crop the data for t_range using DOLfYN s subset method creates a copy t_range_inds t_range 0 lt dat_raw mpltime amp dat_raw mpltime lt t_range 1 dat dat_raw subset t_range_inds Then clean the file using the Goring Nikora method 30 adv clean GN2002 dat dat cin dat copy A Create a figure for comparing screened data to the original fig plt figure 1l figsize 8 4 EGS EE ax fig add_axes 14 14 8 74 Plot the raw unscreened data ax plot dat_raw mpltime dat_raw u Plot the screened data ax plot dat mplitime datk u kOm bads np abs dat u dat_raw u t_range_inds ax text Ona UE 0 2f of the data were cleaned Anby the Goring Nikora method np float sum bads gt 0 transform ax transAxes va top ha left Add some annotations ax axvspan dt date2num dt datetim _range 0 zorder 10 facecolor 0 9 ax text 0 13 0 9 Mooring falling ntoward_seafloor ha center va top transform ax transAxes E len bads B dgecolor none size small ax text t_range 0 0 0001 0 6 Mooring on seafloor T size small ha left ax annotate t_range 0 0 006 rangelol r O23 arrowprops dict facecolor black ha right Finalize the figure
8. because the turbulence measurements were made at dif ferent sites In each panel the doppler noise level arrow points at doppler noise that exceeds the high frequency turbulence levels ee Figure 9 Spectra of turbulence highlighting motion correction A Streamwise velocity B cross stream velocity C vertical velocity Black lines show the uncorrected spectra red show spectra of head mo tion and blue shows the spectra after motion correction Green shading highlights locations where mo tion correction reduced the spectral amplitude The u and v spectra have sharp peaks at 0 1 Hz and lesser peaks at higher frequencies that match the spectrum of i indicating motion contamination Motion correction removed the vast majority of the contamination from these spectra but contamination from the large peak in the v spectra at 0 1Hz persists The w spectra is essentially uncontaminated by the mooring motion The inertial sub range is shaded as it was in Figure8 Figure 10 Spatial coherence estimates from TTMs Vertical coherence estimates A are from ADVs on TTM spaced 0 6m apart Lateral coherence estimates B are between neighboring TTMs spaced 50m apart Dashed lines in both figures indicate the 95 confidence level above which the coherence esti mates are statistically differentfrom0 0 2 0 0 0 000000000000 vi Figure 11 The circuit board and pressure case end cap of a Nortek Vect
9. example more than 30 of the measurements had a U in the range of 0 8 1 2m s 4 2 Turbulence Spectra The primary purpose for making ADV measurements at HKT sites is to measure the turbulence spectra That is ADVs resolve the inflow at a level of detail that cannot be measured with profiling instruments Turbulence spectra are estimates of the distribution of energy as a function of frequency eddy size Because spectra reveal detailed information about the signal velocity they also reveal detailed sources of error in the measurement It is therefore important to be aware of these errors so that one can be careful to exclude them from estimates of statistics meaning ful to the flow Kolmogorov s theory of locally isotropic turbulence predicted that turbulence spectra would have an inertial sub range in which the amplitude of the spectral components i e S u S v and S w will be equal and in which the spectra will decay as KP Kolmogorov 1941 This prediction has been confirmed by observation so ubiqui tously in oceanic and atmospheric turbulence that it has become a defining characteristic of turbulence spectra e g Figure 8A data from Thomson et al 2012 Based on this we expect that deviations from this behavior are likely to indicate some source of error 15Equivalent expressions apply for the v and w components 14 Velocity m s Turbulent Energy m s Reynold s stresses m s 00 00
10. ha center tobox dict facecolor w alpha 0 9 edgecolor none zorder 20 fig2 savefig fig motion_vel_spec pdf 33 3 Data cropping and cleaning 0 41 of the data were cleaned by the Goring Nikora method Mooring falling toward seafloor Mooring on seafloor u m s 12 30 12 20 12 25 12 05 12 10 12 15 Time June 12 2012 Figure 13 The crop_data pdf figure generated by the adv_example01 py script The uncropped uncleaned data is in red and the cropped and cleaned data is in blue 34 Velocity Spectra motion corrected no motion correction 10 Motion Correction m s7 hz frequency hz Figure 14 The motion_vel_spec pdf figure generated by the adv_example01 py script Spikes in the spectra due to motion contamination red are removed by motion correction blue 35
11. of turbulent inflow to a MHKT The mean flow profile is indicated in blue and turbulent eddies of different sizes lengths L and orientations are indicated in red motion sensors IMUs into ADVs which measuring ADV mooring motion so that it can be removed from the measured velocity Section 2 describes hardware details of the mooring used and ADV configuration details specific to moored measure ments Section 3 details the processing steps for transforming moored ADV measurements into earth frame velocity signals and computing turbulence statistics from those measurements Section 4 describes turbulence analysis meth ods that are useful to the HKT industry and defines the applicability and limitations of the results The reader is also encouraged to download and install the Doppler Oceanography Library for pYthoN software package DOLfYN lkilcher github io dolfyn which provides example instrument configuration files and functions for performing the data processing and analysis steps described herein 2 Measuring turbulence Acoustic Doppler velocimeters ADVs are capable of high accuracy precision lt 1 and sample rates up to 64hz These characteristics make them the preferred tool for measuring the spectrum and spatial coherence of turbulence at tidal energy sites However they measure the water velocity within a few inches of the sensor head Therefore in order to resolve turbulence that is relevant to MHKTs these instrum
12. persist This is a notable but relatively minor level of uncertainty in the context of the highly energetic flows that exist at tidal energy sites 29 C DOLfYN data processing scripts The following script details data processing steps adv_example01 py To get started first import the DOLfYN adv advanced programming interface API from dolfyn adv import api as adv Import matplotlib tools for plotting the data from matplotlib import pyplot as plt from matplotlib import dates as dt import numpy as np FEE TE AE HE TE AE HE TE FE HE TE FE HE TE FE HE TE EE EEE EEE HE HE t User input and customization the ile to load fname data vector_data_imu0l vec This is the vector from the ADV head to the body frame in meters in the ADV coordinate system body2head_vec np array 0 48 0 07 0 27 This is the orientation matrix of the ADV head relative to the body In this case the head was aligned with the body so it is the identity matrix body2head_rotmat np eye 3 The time range of interest t_range The instrument was in place starting at 12 08 30 on June 12 2012 ditndate2 numa datetime datetime 2012r er UL SESION The data is good to the end of the file pean This as the filter to use for motion correction accel filter 0 1 End user input section AE AE aE AE aE aT AE AE AE aE eae aE AE AE AE aE EE aE aE Read a file containing
13. stresses the tur bulence spectrum and spatial coherence all contribute significantly to fatigues loads If turbulence is conceptualized as a mixture of eddies of different sizes orientations and rotation speeds Figure 1 the importance of these statistics can be understood as follows e Mean shear can impart a torque on the rotor shaft and induce variable loads on the blades as they rotate through the spatially non uniform mean flow The Reynold s stresses u v u w and v w indicate the orientation of the eddies in the flow Eddies of differ ent orientations may impart forces on different components of the turbine differently For example an eddy aligned with the rotor v w in Figure 1 will impart a large torque on the rotor that eddies of other orientations would not The turbulence spectra quantifies the energy of eddies of different frequencies from which length scales 6 can be estimated For example an eddy with 6 similar to the blade cord keora is likely to impart larger fatigue loads on the blade than a smaller or larger eddy with the same energy Likewise an eddy the same size as the rotor will impart a larger load on the rotor than a much smaller eddy Quantifying the energy in these eddies is therefore important to accurately estimating the loads they induce Spatial coherence quantifies the correlation of the turbulence in space i e the length L of the eddies It is important because longer eddies are likel
14. 3D isotropic turbulence these length scales should be similar For the largest eddies which are expected to be depth limited and thus 2D anisotropic it is likely that L will greatly exceed 6 Knowledge of the length of these large eddies is important to HKT design because they are the most energetic and if their dimensions match that of HKT components they are likely to have a larger impact on the HKT 10In many formalisms of turbulence the Reynold s stresses are components of off diagonal elements of the Reynold s stress tensor In these arenas the diagonal elements of that tensor are the components of Erze Based on the NWTC s FAST wind turbine simulation tool Note that removing a linear trend means that F u F w 13 The u component spatial coherence is estimated as 3 Fu F 1 Si u Si u Where denotes an ensemble average and i and j denote different measurement points in space Tij u f 4 Data analysis Data analysis is the process of synthesizing useful knowledge or information from a set of data The details of data analysis depend entirely on the goals of the analysis and the data available That is what question is attempting to be answered and how suitable is the dataset to answering that question This section presents example analyses of moored ADV data that provides potentially useful information for HKT site and device developers and discussing the accuracy and limitations of the approa
15. Measuring turbulence from moored acoustic Doppler velocimeters A manual to quantifying inflow at tidal energy sites Levi Kilcher Jim Thomson Joe Talbert and Alex DeKlerk Wednesday 3 December 2014 Acknowledgments The authors would like to thank Capt Andy Reay Ellers for his patient and precise ship operations 111 Executive Summary This manual details a set of methods for measuring and quantifying turbulence at tidal power sites It is written to aid site and device developers in quantifying the turbulence statistics that are important to tidal energy converter power performance and lifetime key parameters needed to estimate cost of energy This manual details mooring design instrument configuration data processing steps and analysis guidelines for estimating turbulence statistics at tidal energy sites This provides the tidal energy industry with a low cost methodology for quantifying the turbulent inflow that reduces the operational lifetime of tidal energy turbines This will help the industry to design tidal energy devices that are more reliable less expensive and more efficient which will lower the cost of tidal energy 1v Table of Contents 1 Introduction s s ese eega e 2 k aa a E aa ee 1 2 Measuring turbulence ouou 2 2 1 Mooring Hardware s s acka ea ee 2 2 2 Instrument configuration sses s sssr ssrt ee 3 2 2 1 Record position and orientation of the ADV head oona o o 4 2 2 2 Softw
16. OLfYN s adv io rotate CorrectMotion class To do this you will need to specify H and Tai as properties of your raw cleaned adv data object and select a value for fa e g lines 18 23 34 108 111 of appendix C For those unfamiliar with Python the motcorrect_vectory py script bundled with DOLfYN provides a command line interface for performing this motion correction and saves the motion corrected data in Matlab format In that case H and a are specified in an input orient file and fa can be specified as a command line option 3 2 1 Select a local coordinate system Prior to performing any averaging and computing other statistics it is often useful to rotate the measurements into a locally meaningful coordinate system For the purposes of quantifying turbulence at HKT sites it is common practice to rotate the data into a coordinate system in which u is the streamwise velocity v is the cross stream velocity and w is in the up direction For details on selecting estimating and transforming into such a coordinate system see appendix A 2 3 3 Cleaning data Data cleaning is a two step process of 1 identifying erroneous bad points in an otherwise good dataset and 2 replacing them with either a reasonable estimates of the values at those points or b error values e g NaN not a number which explicitly indicate the points are invalid Several methods exist for identifying bad data T
17. aning against the anchor stack railroad wheels White fair wrap on the mooring lines is used to reduce strumming Plastic zip ties are used to fasten the ADV cables to the vane NMSS u bolts and rubber gaskets fasten the ADV body to the fin A piece of 1 angle NMSS fastened to the backbone is used to extend the ADV heads approximately 10 foreward of the vane s leading edge 3 The body orientation matrix R provides the orientation of the ADV body relative to the earth It is used with H to estimate the earth frame orientation of the ADV head and thus the velocity vector in the earth frame It is also used to remove gravity from the linear acceleration measurement see section 3 The Nortek Vector can be purchased with a Microstrain GDM GX3 25 miniature attitude heading reference system MicroStrain The 3DM GX3 25 can output all three of these channels Sampled and stored in realtime by the same controller the Vector velocity measurements and 3DM GX3 25 motion and orientation measurements are tightly synchronized to within 107 sec allowing for high fidelity motion correction in post processing New versions of the Nortek software bundled with a new Vector allow the user to select which datastreams from the 3DM GX3 25 are stored in the Vector output data file To set a Vector to record the correct data open the Vector program and go to Deployment Planning Use Existing For most situations you should on
18. are configuration ooo e ee 4 2 3 Deployment Planning o ss e esas ee eras ee e E E E 8 3 Data prOCessIME a ose akoa a a e be 8 3d Reading dala 2245 46 254 0 405 23S A eee ASS a ide SESS 9 3 2 Moton Correction p es ap oe ek sd ER ee 6 ROS Be eS Be es 9 3 2 1 Selecta local coordinate system 2 2 2 ee 10 3 3 Cleaning data lor e oe ee eo ee SR PAE RE ee bee Be Ee Ee He 10 3 4 Turbulence metrics and averaging 2 2 0 000 ee ee ee 11 3 4 1 Turbulence intensity lt s e s s coso 002 0 2000000000000 00000 12 3 4 2 Turbulent kinetic energy ooa e 12 3 4 3 Reynoldis Stresses oos ss oce a a SD SEAR a EERE ORE RE ES SS 13 3 4 4 Turbulence auto spectra o oo ee 13 345 Spatial coherence s 2 55 4 en rd a E a da 13 4 Dataanalysis 6d as AA AA i eae eS 14 4 1 Initial inspection time series and histograms e 14 4 2 Turbulence Spectra ea ee a a ee A 14 4 3 Spatial Coherence se seo de eaga RE RR RE Ee ee ee 17 5 SUMMA a Baas he SG Ba Se REAR SAD ES RES OA pe BARE Dade de RS e 18 References 35 ri bb SAE ee SEE EEE RNS ESE SL ee SEE PSS 21 A Coordinate systems ee 23 A l Defining coordinate systems ooo 23 A 2 Stationary frames e ce t ee 23 A2 1 The earth frame po sepoi 5554 Se RRR Re RSS ada ee a 23 Ac22 The analySISTTAME s goce a A A ee Ge pe eee YA E 24 A3 Measurement frames y oat cg bs ee PRE OEE Aw HAE SOY eR RS ESSE os ERED 24 A31 The ADV head a
19. ce code for details DOLfYN also adds l u to feag to estimate hea The orientation matrix In order to use the orientation matrix to rotate velocity measurements into an earth fixed coordinate system it is important to understand how the orientation matrix is defined The Microstrain IMU outputs an orientation matrix Rimu such that gt imu __ gt NED u Rimu Y Where i and NED are vectors in the IMU s local coordinate system and a north east down NED earth fixed coordinate system respectively MicroStrain 2012 However this NED earth coordinate system is different from the ENU earth coordinate system used here and typically used by Nortek R stad 2011 That is i B NEP 26 View direction Y X Figure 12 Coordinate systems of the ADV body and head A A strongback with an ADV rests on a block of wood Coordinate systems of the ADV head magenta and body yellow are shown The ed direction is known by the black band around the transducer arm and the x direction is marked by a notch on the end cap indiscernible in the image The cyan arrow indicates the body to head vector It g The perspective slightly distorts the fact that lt 4 2 shed 5 and zhead 3 B Coordinate system of the ADV head as defined in the Nortek Vector manual Nortek 2005 27 where 0 1 0 B 1 0 O 0 0 1 From this and the above discussion of the orientation of the IMU in the ADV it i
20. ce frames or simply frames is a critical and somewhat tedious task for making accurate velocity measurements using moored ADVs The coordinate systems for doing so can be broken into two categories 1 the inertial or stationary ones into which it is the goal to transform the measurements and 2 the moving coordinate systems in which sensors make measurements The purpose of this appendix is to clearly document and define the relationships between all of the coordinate systems necessary for quantifying turbulence using moored ADVs This appendix starts with general definitions of coordinate systems and the relationships between them A 1 then details the stationary and measurement frames used herein A 2 and A 3 respectively A 1 Defining coordinate systems Consider two three dimensional right handed coordinate systems a and b with orthogonal basis vectors 2 and 2 In general these coordinate systems are related by the equation P Ri R Here superscripts denote the coordinate system that the quantity is measured in and indicates standard matrix multiplication The vectors 7 and 7 point to the same point in space but in the two distinct coordinate systems In this framework the vector is the translation vector that specifies the origin of coordinate system b in the a frame and R is the orientation matrix of b in a With these definitions the foll
21. ch 4 1 Initial inspection time series and histograms As a first step in most analysis of velocity data it is useful to plot the velocity and other turbulence statistics as a function of time In the example data in Figure 6 the tidal currents reach 2m s During this period at this location the floods are significantly larger than the ebb The mean velocity appears to be a reasonable estimate there is a clear tidal signal there are no sudden dramatic jumps in the values and the magnitude of the velocity agrees with previous measurements at this site this gives us confidence that our methods have produced a reliable dataset Figure 6A The instantaneous turbulence intensity has an average of 10 but approaches 20 in some 5 min periods 6B As is often observed in turbulent flows throughout the oceans and atmosphere the turbulence is highly intermittent that is it is dominated by large spikes and periods of relative calm 6C Note also that the turbulence is significantly lower for the small ebb than it is for the two larger floods The Reynold s stresses show a similar pattern 6D HKT site developers often use histograms of velocity measurements to estimate the available power at a tidal energy site The record in Figure 6 isn t long enough to estimate annual energy production or AEP but a histogram of the measurements does provide some indication of the distribution of velocity at the site Figure 7 During this time period for
22. contaminate the results The wind energy industry uses At 10min Commission 2005 This is appropriate for large modern wind turbines with long ramp up times but the smaller size of current HKTs suggests that they may respond faster and therefore that a smaller At might be appropriate Gunawan et al 2014 On the other hand if turbulence is to be treated as the primary driver of device fatigue loads one should be careful not to implicitly neglect energetic low frequency turbulence by selecting A to be too short With these considerations in mind and until further work provides details on the relationship between turbulent inflow and HKT loads this document recommends using At 2 10 minutes The exact choice of At within this range is unlikely to alter the results significantly and should be adjusted depending on the goals of the analysis For example when fitting theoretical spectra to observations for the purpose of input to stochastic flow simulation tools such as TurbSim Jonkman 2009 it is desirable to include lower frequencies in the fit and therefore it is reasonable to use a longer At 5 10min With A chosen the ADV data record is broken into segments in which turbulence statistics are computed In this way the time series of instantaneous velocity d at the instrument sample rate e g 16Hz is converted to a time series of turbulence statistics with time step Ar Figure 5 It is recommended to save your data at this l
23. cra ag p58 pa a doe fob dee Se ee ee ae e bd 25 A 3 2 The IMU coordinate system 2 ee 26 B Filtering Acceleration haaa aa 29 C DOLfYN data processing scripts 2 2 30 List of Figures Figure 1 Diagram of turbulent inflow to a MHKT The mean flow profile is indicated in blue and turbu lent eddies of different sizes 6 lengths L and orientations are indicated inred Figure 2 Schematic diagram of the tidal turbulence mooring TTM o o Figure 3 ADVs mounted on a strongback vane prior to deployment The heads and bodies are tilted at 15 to account for mooring blow down NMSS shackles and pear links connect the strongback to the moor ing lines The strongback is leaning against the anchor stack railroad wheels White fair wrap on the mooring lines is used to reduce strumming Plastic zip ties are used to fasten the ADV cables to the vane NMSS u bolts and rubber gaskets fasten the ADV body to the fin A piece of 1 angle NMSS fas tened to the backbone is used to extend the ADV heads approximately 10 foreward of the vane s leading COBO gris cc cs eh ic aie eee eee ease Ge pelos A tdi do Roses soon a A ERR ae ee Ge Figure 4 The Nortek Vector program s deployment planning pane with some typical settings for quantify ing turbulence at tidal energy sites Required settings are highlighted in red and recommended settings are in green The blue box points out the bat
24. data record 3 Seal pressure cases carefully and install dessicant moisture absorbing packs to reduce the risk of water damage to electrical hardware 4 Synchronize instrument clocks to a single computer clock that has been recently syncronized to internet time via Network Time Protocol NTP 5 Configure the instrument appropriately for the deployment Perform bench tests several weeks prior to the deployment to allow time to replace faulty components if necessary lt q _ 37 Steel float 700Ibs Bouyancy 5 8 SAS 3 Ton Esmet Swivel _ ___ Amsteel Line A 5 8 SS Shackle lt Nortek IMU ADV on strongback lt 5 8 SS Shackle 4 Y Amsteel Line A 5 8 SAS lt 4 3Ton Esmet Swivel lt _ 5 8 SAS 4 ORE 8242 Acoustic Release lt t ___ 5 8 SAS Drop Link 4 Y Galv Chain lt ___ 5 8 SAS 3 RR Wheel Anchor Stack 2500Ibs wet Figure 2 Schematic diagram of the tidal turbulence mooring TTM In order to produce high fidelity spectra and spatial coherence estimates from moored ADV measurements motion sensor measurements must be tightly synchronized with ADV velocity measurements Currently Nortek s Vector Mis the only instrument that can be purchased off the shelf with a tightly synchronized inertial motion sensor IMU These instruments were used for the TTM te
25. e 10 00 00 04 00 08 00 12 00 Time June 16 2014 0 50 100 150 200 250 300 Time seconds Figure 5 An example velocity time series measured using a TTM at Admiralty Inlet A The mean stream wise velocity i blue Ar 5min is over layed on the full signal grey B A 5min data window of the turbulent piece of the streamwise velocity u The dashed lines indicate one standard deviation After points have been identified as bad they will need to be replaced For cases of a small number of sparsely distributed bad data points they can be replaced with interpolated values DOLfYN s adv clean GN2002 function uses a least squares cubic polynomial without introducing significant interpolation related bias This approach produces a dataset that can be further processed without the headache of dealing with NaN values If on the other hand there are segments of data with large fractions of bad points gt 10 20 interpolation may introduce significant bias to a myriad of statistics of the data In these cases it is best to crop out the bad segment perhaps creating two distinct data sets that can be rejoined later or to assign NaN values to the bad points In general the choice between these options will depend on the objectives of the analysis or the preference of the inves tigator For the purposes of this document in which spectral analysis is a primary result assigning NaN values will make spectral analysis difficult to the po
26. e is controlled by the outer scale of the forcing That is the lateral spatial coherence is con trolled by the distance from the bottom This indicates that turbulent loading on devices in an array with hub heights smaller than their separation distance will be uncorrelated across the array 5 Summary This document outlines the methods for making turbulence measurements at HKT sites using mooring deployed IMU equipped ADVs The critical issue addressed by this approach is that ADVs which accurately measure turbu lence statistics are deployed at heights above the sea bed that are relevant to HKTs Other existing approaches most of which deploy instrumentation on the seafloor either ADPs or ADVs do not resolve the statistics of the turbu lence with sufficient accuracy ADPs or at the correct location ADVs to produce reliable device lifetime estimates This manual provides guidance on 1 designing mooring hardware that can support the instrumentation 2 planning deployment and recovery to capture the statistics of turbulence that are important to HKTs 3 configuring instru mentation for data collection 4 processing data and 5 tips for analyzing data to produce useful results It is highly recommended that users of this manual also download and install the DOLFYN software package as each data pro cessing step described herein can be performed in a few lines of code In particular the tedious details of accounting for different coordi
27. ents must be positioned near the hub height of the MHKTs that will be deployed at that location This section describes the mooring used and the details the configuration of instrumentation for measuring turbulence from a moored platform 2 1 Mooring Hardware The Tidal Turbulence Mooring TTM mooring is a simple compliant sub surface mooring Thomson et al 2013 Its primary components are a clump weight style anchor an acoustically triggered release for mooring re covery mooring lines a strongback vane at turbine hub height that orients the ADVs into the flow i e passive yaw and a buoy that holds the mooring lines taught Figure 2 The clump weight is composed of 3 railroad wheels stacked on a central steel cylinder A steel flange welded securely to the base of the cylinder supports the weight of the wheels The total wet weight of the anchor is 2500lbs Galvanized 1 2 anchor chains and a 5 8 steel shackle connect the top of the anchor stack to the acoustic release manufacturer ORE now EdgeTech At the top of the acoustic release a high tension swivel allows the mooring line to rotate without imparting torque on the hardware below The buoy for the TTM is a 37 diameter spherical steel buoy manufacturer McClane that is pressure rated for the depths to which it will be deployed Another high tension swivel between the buoy and mooring line allows the buoy to spin without imparting large torques on the mooring
28. eps and for saving data along the way See appendix C for an example processing script 3 1 Reading data ADVs typically record data internally in a compact vendor defined binary format The vendor will generally publish the details of the data format e g R stad 2011 and also release software tools for viewing this data and or writ ing it to other common data formats e g white space comma tab delimited formats Matlab format or other increasingly common standards such as HDF5 Nortek provides software tools for converting raw binary vec files to Matlab format see http www nortek as com en support software and the DOLfYN software package is capable of reading these files directly into Python NumPy arrays appendix C line 40 A dataset will also generally need to be cropped to the period of interest e g when the instrument was in place on the seafloor Figure 13 3 2 Motion Correction Raw turbulence measurements from moored ADVs will be contaminated by mooring motion When IMU mea surements are tightly synchronized with standard ADV velocity estimates the IMU measurements can be used to reduce this contamination This involves removing the measured velocity induced by ADV head motion ti from the measured velocity u to estimate the motion corrected velocity in the earth frame W t thn t H t 1 Here superscript e s denotes the earth coordinate system We now break ui into tw
29. er comparison of the velocity spectra to the spectra of uncorrected velocity measurements S i and spectra of the head motion S i shows that this peak is indeed due to mooring motion Figure 9 This comparison is remarkable because it highlights the effectiveness of the motion correction method At many frequencies head motion is 5x larger than the corrected signal which is believed to be correct because it agrees with a f e spectral slope Furthermore with the understanding of isotropy and the f 5 3 slope there is a strong theoretical footing to simply interpolate over the peak in S v to estimate the underlying real spectrum These results suggest that motion corrected moored ADV measurements are capable of producing accurate estimates of all three components of the turbulence spectra 4 3 Spatial Coherence Spatial coherence is the highest order turbulence statistic that is considered in this document As such it is highly sensitive to the details of the inflow and measurement method Methods for measuring this variable over the scales important to HKTs e g rotor diameter permit inflow simulations that accurately resolve the spatial correlations of turbulence at HKT sites Vertical spatial coherence estimates from two IMU equipped ADVs deployed on a TTM separated by 0 6m are plotted in Figure 10A The lack of motion contamination in S wm Figure 9 suggests that this component will have an accurate estimate of Ta w Indeed t
30. etermine hydro kinetic turbine HKT lifetime Device simulation tools such as HydroFAST and Tidal Bladed have been developed to estimate HKT power performance and lifetimes based on device mechanical electrical models and realistic inflow conditions These tools help to accelerate the HKT industry by 1 helping device designers predict failure modes in prelimi nary designs and 2 providing site developers and financial institutions with estimates of the cost of energy for a particular device Device designers typically have all of the information necessary for producing device models for these simulations but often lack adequate knowledge of turbulent inflow conditions to produce accurate power performance and life time estimates This gap arises from the difficulty and high cost of making turbulent inflow measurements that are relevant to HKTs Furthermore the method for measuring turbulent inflow must be suited to the energetic sites where HKTs will be deployed This work aims to fill this gap by detailing a relatively low cost and robust methodol ogy for measuring the turbulent inflow statistics relevant to HKTs The question What turbulent statistics determine turbine device performance and fatigue loads motivates an ac tive area of research in both the wind turbine and HKT industries While no single statistic or group of them has been identified that fully predicts fatigue loads there is broad agreement that mean shear Reynold s
31. evel to allow quick and easy access during analysis The remainder of this section defines several turbulence variables that are commonly used in the HKT industry and can be computed from moored ADV measurements Furthermore DOLfYN s adv io turbulence TurbBinner provides a two line interface for performing averaging and computing all of these statistics e g lines 124 125 in appendix C 3 4 1 Turbulence intensity Turbulence intensity is used throughout the wind industry and other engineering fields as a zeroth order metric for quantifying turbulence It is defined as the ratio of the standard deviation of horizontal velocity magnitude U Vu v to its mean _ std U o OU I Turbulence intensity is often quoted in units of percent i e 100 J It is useful because it is easy to understand and in many observations of atmospheric and oceanic turbulence is relatively constant for U gt 0 On the other hand J has often been criticized for being too simple in particular that it only includes information about horizontal velocity such that it does not provide enough information about the turbulence for various applications 3 4 2 Turbulent kinetic energy Some discussions include a wave velocity but for simplicity it is not included here at this time 12 Turbulent kinetic energy tke quantifies the total energy contained in turbulence Eike ulul ww Like J Eike is useful because it is relatively simple As
32. he shape of I 4 w agrees with measurements of coherence from 14White noise has constant amplitude with frequency 17 Vertical spatial coherence Lateral spatial coherence Az 0 6m TTM1 Ay 50m TTM1 to TTM2 1 0 B Successful Motion 0 6 Correction 0 4 Persistent Motion Contamination 0 21 10 10 10 f Hz f Hz Figure 10 Spatial coherence estimates from TTMs Vertical coherence estimates A are from ADVs on TTM1 spaced 0 6m apart Lateral coherence estimates B are between neighboring TTMs spaced 50m apart Dashed lines in both figures indicate the 95 con fidence level above which the coherence estimates are statistically different from 0 other environments i e it has an exponential decay Kilcher et al 2014 Unfortunately 4 u and y v are contaminated by persistent mooring motion contamination at 0 1Hz The peak in coherence arises because the measurements were made from the same TTM strongback vane and the co motion of that vane created a peak in the coherence estimate at the frequency of mooring motion The lateral spatial coherence estimates between ADVs on two different TTMs shows zero spatial coherence We believe this to be a reliable and important result at this site SOm water depth and distance above the bottom 11m turbulence is incoherent over spatial separations of 50m This supports the theory that the limiting scale for spatial coherence of turbulenc
33. hese include 1 Search the data for manufacturer defined error values if the manufacturer defines these 2 Search for values outside a reasonable range For example tidal velocities are typically less than 4m s there fore a velocity measurement greater than 5m s is probably bad Histograms can be useful for identifying the reasonable velocity range The distribution of velocity measurements will often be approximately Gaussian values well beyond the tails of the distribution gt 3 standard deviations can probably be identified as bad 3 Utilize diagnostic data from the instrument to identify bad data For example low values of correlation the similarity of the send and receive acoustic pulses can sometimes indicate bad data 4 Apply spike detection algorithms to the velocity signal While turbulence is by definition unsteady and chaotic it is not discontinuous Large and sharp spikes in the velocity signal are almost always bad Modern ADVs often produce data of sufficiently high quality that only a relatively small number of spike type bad data points are present in the raw data after cropping In these cases spike detection is usually sufficient to identify bad points The recommended spike identification method is documented in Goring and Nikora 2002 and Wahl 2003 and implemented in DOLfYN s adv clean GN2002 function The velocity data in Nortek Vector vec files do not contain an error valu
34. int that it is best to simply split the data record to remove the bad segments and plan to recombine in later stages of processing 3 4 Turbulence metrics and averaging Having cleaned the raw data and computed an estimate of the velocity vector in a useful coordinate system ti one can finally begin estimating the turbulence statistics and average mean flow properties the measurements were de signed to capture For each component of velocity i u v w turbulence is defined by separating the instantaneous 11 velocity e g u into average and turbulent u pieces u atu a 4 Where the over bar denotes a suitable average over a period Af such that w 0 For HKTs it is useful to choose At such that i is the flow which the turbine is designed to efficiently convert into useful energy while the turbulence is what contributes to fatigue loads that decrease device lifetime That is At should be somewhat longer than a typical HKT ramp up time tens of seconds to a minute or two Defined this way a turbine can be considered to be in a steady operational state so long as changes in are small compared to i itself Turbulent velocity fluctuations can then be treated as disturbances to a HKT s steady operation A time scale At must be chosen such that the tidal flow has stationary statistics i e stable mean and variance for that duration of tide If Ar is too long the tidal variation itself will
35. ity 7 Set Speed of Sound to Measured The Nortek Vector uses a temperature sensor and a fixed Salinity value to calculate an estimate of the speed of sound which is important to the velocity measurements If the salinity of your site is not known consider measuring it 8 Modify the burst interval settings to maximize data return If you are going to the effort of putting this instru ment in the water it is valuable to capture as much turbulence data as possible Once the above settings have been set follow these guidelines to maximize data recovery A Use Continuous sampling if possible Consider purchasing additional batteries With the settings recommended above 2 Lithium batteries will last 11 days and 2 Alkaline batteries will last 3 3 days 3The magnetometer signal is not needed for motion correction but the other three signals are Note that this is the only option that provides the orientation matrix which is required for motion correction Standard Advanced Setup Speed of sound Deployment planning Sampling rate 9 Measured Battery pack 2 Alkaline zi i Salini i 35 Nominal are Battery capacity Wh 100 velocity range 1525 Fixed m s Assumed duration days 3 9 Continuous samplin Battery utilization vn Geography of capacity Burst interval s 600 Memory required MB 442 Number of O a i samples per burst 10 jia Vertical vel range m s 15 Comino ayuno gt Surfzone Horizontal
36. line Half inch Amsteel line e g http www amsteelblue com is used to connect the strongback vane to the buoy and acoustic release using 5 8 shackles Amsteel line has a high strength to weight ratio low stretch and low torque The half inch line used here has a breaking strength of 30 600lbs much larger than the dry weight of the mooring lt 3 500lbs If modifications are made to the mooring design to avoid catastrophic damage be sure that the mooring line can safely support the weight of the mooring during recovery a safety factor of at least 5x is recommended The blow down angle of the TTM was simulated using University of Victoria s Mooring Design and Dynamics software The observed blow down angle of 20 at 2m s agreed well with the predictions Thomson et al 2012 This mooring design has been safely deployed in currents up to 3m s without exceeding a maximum advisable blow down angle of 40 If significant modifications are made to this mooring design such as changes in mooring length deployment depth or other modifications to major hardware components or if operating in much stronger currents the new design should be re simulated using a mooring simulation tool to determine blow down angle as well as tension and drag forces The strongback vane was designed to be a robust and low cost component that effectively holds an ADV head or two upstream of the mooring line and holds the ADV body nearby and rigidly fi
37. ly need to use the Standard tab The following settings are required to be able to perform motion correction Figure 4 1 Check the box to the left of IMU This tells the Vector to use and record information from the IMU 2 Select Accl AngR Mag xF from the drop menu to the right of IMU This tells the Vector to record the acceleration Accl Angular Rate AngR Magnetometer Mag and Orien tation Matrix xF signals 3 For the Coordinate system select XYZ This instructs the Vector to record data in the ADV head coordinate system For most measurements of turbulence at tidal energy sites the following recommendations are also likely to be appropriate 4 Setthe Sampling Rate to 16 Hz In most tidal environments lower sampling frequencies will not resolve all of the turbulence scales Higher sampling frequencies are typically dominated by instrument noise 5 Set Geography to Open ocean This instructs the ADV to operate in high power mode which increases data quality Consider upgrading batteries use 2 Lithium batteries if necessary or using burst sampling before using lower power Surf zone or River setting See the instrument manual for further details 6 Setthe Nominal velocity range to 4 m s Tidal velocities at most tidal energy sites will be in this range Use the higher range of 7 m s if you have reason to believe the velocities will be larger than 4 m s at the expense of some data qual
38. mate provided by the Nortek Vector software is as accurate as possible 2 3 Deployment Planning Safe efficient and accurate deployment of scientific equipment in the oceanic environment is a science of its own Carefully considered planning is critical to deployment safety and success HKT sites are generally locations of strong currents that add to the difficulty and complexity of deploying scientific equipment It is highly recommended that deployment in these environments be led by experienced professionals in the field At the very least enlist such professionals to advise your planning and deployment process A well written research cruise plan should include the following information e Scientific objectives A detailed schedule of all activities needed to accomplish objectives Noteworthy environmental conditions at the deployment site e g tidal amplitude current amplitude probable weather conditions daylight hours etc Schematic diagrams of hardware that will be deployed A list of all personnel involved in the deployment Maps of deployment locations that include notable bathymetric features e g sub surface ridges or canyons and human infrastructure hazards e g buoys or other equipment e A risk assessment and risk mitigation plan The schedule is one of the most important parts of the cruise plan It should include personnel arrival departure times ship arrival departure times ship loading and ship prepa
39. me scales can be converted into length scales using Taylor s frozen flow hypothesis i e l fi spectra quantify the distribution of turbulent energy at different length scales When considering turbulence to be a complex interaction of eddies from very small to very large scales the spectrum quantifies the energy rotation speed of the eddies as a function of their size 6 Figure 1 HKTs respond to different scales of velocity fluctuations different eddy sizes differently HKT simulation tools such as Tidal Bladed and HydroFAST are capable of estimating the loads induced by these fluctuations but the critical information of how energetic those fluctuations are must be provided as input to these tools Fortunately spectra provide exactly this information the distribution of energy as a function of eddy size Spectra are estimated from Fourier transforms Fast Fourier Transforms of the turbulent velocity S u f F 1 In this work FFTs denoted by are computed by removing a linear trend fit from u and applying hanning windows to reduce spectral reddening Priestley 1981 Spectra are normalized so that S u f df ww 3 4 5 Spatial coherence Spatial coherence is an estimate of the correlation of velocity components over spatial distances as a function of frequency That is where spectra indicate the energy in eddies as a function of their size 6 coherence is an estimator of their length L For
40. n defined by the right hand rule relative to and 2 and 2 is the vertical up direction Note that throughout this work vector quantities with no superscript are in this local frame The orientation of this frame relative to the earth is defined as cos sing 0 S sin coso 0 0 0 1 Where 6 is the angle from East to the streamwise direction For the purpose of estimating it is convenient to use complex notation for the horizontal velocity O u iv Uet Where i I e 2 71828 is Euler s number U is the instantaneous horizontal velocity magnitude and is the angle of that velocity from east For measurements at locations where the flow does not change direction dramatically over the measurement period e g in rivers the streamwise direction can be estimated simply from an average of the horizontal velocity over the entire data record Oriver arg O data Where data denotes an average of all data and arg returns the complex angle of its argument For measurements at locations where velocity changes direction over the measurement period a more sophisticated method for determining a local coordinate system is often required For tidal flows for example it is often useful to define the streamwise direction to be parallel with ebb and opposite flood or vice versa This can be done by first defining 29 for 0 lt lt 7 gi 2 2 for pi lt p lt 2x That is angle
41. nate systems have been simplified therein appendix A Finally in section 4 we demonstrate analysis methods for producing estimates of turbulent quantities that are use 18 ful to HKT device and site developers Most importantly the approach outlined in this method produces reliable estimates of the Reynold s stresses turbulent energy and turbulence spectra The w component of vertical spatial coherence can be estimated from a single mooring and the research team behind this manual is developing a new more stable platform for measuring the u and v components of spatial coherence 19 Feedback If you have comments questions suggestions or corrections please contact Levi Kilcher at the National Renewable Energy Laboratory He will be happy to address your concerns if time and resources permit Send comments to Levi Kilcher NWTC 3811 National Renewable Energy Laboratory 1617 Cole Blvd Golden CO 80401 3393 United States of America levi kilcherOnrel gov 1 303 384 7192 20 References Commission LE 2005 Wind Turbines Part 1 Design requirements Egeland M N 2014 Spectral evaluation of motion compensated adv systems for ocean turbulence measurements Ph D Thesis Florida Atlantic University Goring D G Nikora V I 2002 Despiking acoustic doppler velocimeter data Journal of Hydraulic Engineering 128 pp 117 126 Gunawan B Neary V S Colby J 2014 Tidal energy site resource asse
42. ng this file Once you have finished setting your configuration save it to disk for later use Before starting a deployment make sure that your computer s system time has recently been synchronized via NTP and that it is set to the timezone you wish the ADV data to be recorded in If you make changes to your system time you may need to restart your computer to make sure the new time settings propagate to the instrument When your system clock is updated and your configuration is ready consider starting the deployment If the instru ment will not be deployed immediately consider using the delayed start feature to reduce the risk of deploying an 4 At 16Hz a 10min burst will have 10min 60sec min 16samples sec 9616 samples SIt is not recommend to use recharged Li Ion batteries The capacity of these batteries can be less than expected after several uses SOlder versions of Nortek s Vector configuration software will not load this file correctly be sure you have a version that supports the Microstrain chip version 136b6 or later instrument that does not collect data Take care to make sure that the start times agree with the actual physical de ployment times in your cruise plan set the instruments to start before you deploy them so that you can listen to the transmit transducer to confirm that it is operating It is also important to synchronize your computer s clock prior to powering down so that the clock drift esti
43. nt where the head cable meets its end cap The basis vectors of this coordinate system are defined as Figures 11 and 12B points from the center of the head end cap toward the notch in that end cap y is defined by the right hand rule based on the other two basis vectors and 2 points from the head end cap toward the battery end cap along the body cylinder pressure case axis A 3 1 The ADV head In order to transform measured velocities into a meaningful reference frame and to perform motion correction the orientation and position of the ADV head in terms of the body coordinate system must be known To facilitate this the orientation matrix of the ADV head H and translation vector leag are defined according to pu H Y Thea 5 where x 4 and x are the same point in the head and body coordinate systems respectively Combined with the math notes in the previous section the velocity vectors in the head frame can be transformed into the body frame by n HT gread 6 For Nortek Vectors the coordinate system of the ADV head is centered on the transmit transducer face and the coordinate directions are defined by Figure 12B Nortek 2005 16The position of the ADV head origin transmit transducer in the body coordinate system 25 ad the direction of one of the transducer receive arms marked with tape or paint head is defined by the right hand rule based on the other
44. o parts ii i U The first is an estimate of the linear motion of the ADV RO HEO uud 0 Here amp t is the IMU measured acceleration signal rotated into the earth frame and HP f denotes an appropri ate high pass filter of frequency fa The acceleration must be high pass filtered in this way to remove the influence of gravity and that of low frequency bias bias drift that is inherent to IMU acceleration measurements for further details see appendix B and Egeland 2014 Tie amp t R t d t see appendix A for details The second component if uf is due to rotational motion of the ADV head about the IMU T t RT t x i 3 Here is the IMU measured rotation rate a T is the vector from the IMU to the ADV head x indi cates a cross product and superscript xs denotes a quantity in the ADV body coordinate system This coordinate system is used explicitly here to emphasize that is constant in time Matrix multiplication denoted by with the inverse ADV body orientation matrix RT is used to rotate the body frame rotation induced velocity into the earth frame For details on these coordinate systems and the definition of the orientation matrix see appendix A The minus signs in equations 2 and 3 are correct because the measured velocity induced by head motion is in the opposite direction of the head motion itself All of the steps can be performed using D
45. or equipped with a Microstrain IMU The ADV body coordinate system yellow is depicted on the right The notch in the end cap de fines the direction out of the page and the 2 direction points back along the pressure case axis A zoom in on the Microstrain chip highlights its coordinate system magenta relative to the body 25 Figure 12 Coordinate systems of the ADV body and head A A strongback with an ADV rests on a block of wood Coordinate systems of the ADV head magenta and body yellow are shown The 4 direction is known by the black band around the transducer arm and the direction is marked by a notch on the end cap indiscernible in the image The cyan arrow indicates the body to head vector Posa The per spective slightly distorts the fact that head 2 ghead 5 and head B Coordinate system of the ADV head as defined in the Nortek Vector manual Nortek 2005 o o 27 Figure 13 The crop_data pdf figure generated by the adv_example01 py script The uncropped uncleaned data is in red and the cropped and cleaned dataisinblue 0 00 0 34 Figure 14 The motion_vel_spec pdf figure generated by the adv_example01 py script Spikes in the spec tra due to motion contamination red are removed by motion correction blue 35 Vil 1 Introduction Turbulence is a dominant driver of the operational and extreme loads that d
46. owing statements are true e a vector can be mapped from one coordinate system to the other by b __ pa a ue R pu e The inverse rotation is simply the transpose R Rp Ry e The determinant of the rotation matrix is 1 det R 1 A 2 Stationary frames Throughout the main body of this document measurements are discussed in terms of two stationary coordinate systems a the earth frame is the coordinate system in which motion correction is performed and b a local analysis frame coordinate system in which turbulence is analyzed and discussed A 2 1 The earth frame The earth coordinate system is the coordinate system in which the orientation of the ADV is measured see A 3 2 and is the coordinate system in which motion correction is most easily calculated and discussed section 3 2 This work utilizes e superscripts to denote an ENU earth coordinate system with basis vectors 2 East North and 2 Up 15No doubt the earth reference frame is not technically inertial but for the purposes of measuring turbulence in tidal straits we consider it to be so 23 A 2 2 The analysis frame The choice of analysis frame will in general depend on the data available and the goals of the analysis For quanti fying inflow to HKTs it is common practice to use a coordinate system in which X is the streamwise or flood direction is the cross stream directio
47. ration periods transit periods between port and de ployment locations and target deployment times Deployment and recovery should be conducted during slack tides to minimize risks associated with the strong currents present during ebb and flood When preparing the schedule take care to realistically assess the time needed for each piece of the cruise when in doubt allow contingency time for each period The instruments should be deployed for at least two full tidal cycles and ideally up to 6 or more tidal cycles Site characterization measurements for power availability should be much longer see Polagye and Thomson 2013 3 Data processing Data processing involves reading raw data from an instrument and converting it to a form that can be readily ana lyzed in general it includes cleaning removing bad data points converting to consistent scientific units produc ing high or mid level variables from the raw data and averaging Often times during the analysis stage unexpected or unrealistic results will indicate errors in the data When this happens it is necessary to cycle back through data processing steps to inspect raw data and locate the source of the error and clean remove bad data The difference between good and bad data is generally stark but when it is not and no viable justification can be made for removing the unexpected data one should err on the side of keeping the data and if necessary treating it as a special ca
48. s in the lower half of the unit circle are rotated to be in the opposite direction and then all angles are doubled so that fills out the unit circle again When re combined with the velocity magnitudes can be used to estimate the ebb flood direction as at Aide arg Ue ata 2 The ambiguity over whether q points in the direction of ebb or flood can easily be resolved by knowledge of the tidal geographic context of the measurements A 3 Measurement frames To combine signals from an IMU with those of an ADV to perform motion correction the coordinate systems in which each of the measurements are made must be carefully accounted for 24 3DM GX39 25 0EM Attitude Heading Reference System 6223 Y Pe serpin A A602 62 MicroStrain www microstrain com 30M GX3 25 0EM Attitude Heading Reference System 6223 x 4260 11227 2 M www mi Figure 11 The circuit board and pressure case end cap of a Nortek Vector equipped with a Microstrain IMU The ADV body coordinate system yellow is depicted on the right The notch in the end cap de fines the lt direction out of the page and the 2 direction points back along the pressure case axis A zoom in on the Microstrain chip highlights its coordinate system magenta relative to the body For the Nortek Vector instruments that were used for this work the ADV body coordinate system is defined as being centered on the cylinder body axis at the poi
49. s simple to show that the orienta tion matrix of the ADV body in a ENU earth frame is R A Rimu B gt The DOLfYN software package makes this transformation when reading the orientation matrix from Nortek Vector vec files i e the orientmat attribute in the data object returned by DOLfYN s io read_nortek is R not Rimu This way vectors in the body frame can be rotated into the ENU earth frame by i R it 28 B Filtering Acceleration Bench tests of the Microstrain IMU indicate that its accelerometers drift for frequencies lt 10 7Hz a minute or more Egeland 2014 Therefore in order to remove bias drifts in d that when integrated according to 2 lead to large errors in us this document recommends using fa 0 033Hz 30seconds On the other hand real motions at and below fa will not be accurately accounted for in 7 and will therefore persist as low frequency motion contami nation not corrected for in the estimate of ii For moorings whose low frequency motion is limited by the mooring line itself this is a reasonable approach As suming that the displacement of the ADV head from the mooring s neutral position is likely to be lt 20 of its distance from the bottom then for ADVs deployed at 10m depth the speed of their low frequency motion i e be low fa 0 03 Hz will be lt 0 07m s In other words for a 10m mooring the choice of fa 0 03Hz allows for low frequency motion contamination on the order of 7cm s to
50. se Fortunately the reliability of velocity measurements from instruments such as ADVs and ADPs is high the uncer tainties well understood and methods for cleaning data are well defined Instruments generally provide estimates of various sources of uncertainty and other errors as part of their output data streams e g error velocity and beam correlation that aid in cleaning data As the scientific community has become familiar with these measurements well defined and justified methods for cleaning data have been developed and shown to be effective e g Goring and Nikora 2002 This means that it is now possible to generate meaningful and reliable statistics with minimal user input and inspection The DOLfYN software package includes tools and scripts for processing and analyzing turbulence measurements made following the procedures in this document There are four major steps to processing moored ADV data 1 read the raw data from the ADV data file and crop it to the period of interest 2 remove ADV head motion from measured velocity and rotate data into a useful coordinate system 3 clean erroneous points from the ADV data record 4 compute turbulence statistics and averages It is common practice to save the results of these steps some or all of these steps so that later analysis does not require reprocessing which sometimes requires significant CPU time The DOLfYN software package has tools for performing each of these st
51. ssment in the East River tidal strait near Roosevelt Island New York NY USA Renewable Energy 71 pp 509 517 Jonkman B J September 2009 TurbSim user s guide version 1 50 NREL TP 500 46198 National Renewable Energy Laboratory Kelley N Hand M Larwood S McKenna E January 2002 The NREL Large Scale Turbine Inflow and Response Experiment Preliminary Results NREL CP 500 30917 National Renewable Energy Laboratory Kelley N D Jonkman B J Scott G N Bialasiewicz J T Redmond L S August 2005 The impact of coherent turbulence on wind turbine aeroelastic response and its simulation NREL CP 500 38074 National Renewable Energy Laboratory Kilcher L Thomson J Colby J April 2014 Determining the spatial coherence of turbulence at MHK sites 2nd Marine Energy Technology Symposium Seattle WA Kolmogorov A N 1941 Dissipation of Energy in the Locally Isotropic Turbulence Dokl Akad Nauk SSSR 32 1 pp 16 18 URL http links jstor org sici sici 0962 8444 2819910708 29434 3A1890 3C15 3ADOE1IT1L 3E2 0 C0 3B2 C MicroStrain I 3DM GX3 25 Miniature Attitude Heading Reference system product datasheet LORD Micros train URL http files microstrain com 3DM GxX3 25 Attitude Heading Reference System Data Sheet pdf April 2012 3DM GX3 15 25 MIP Data Communications Protocol 459 Hurricane Lane Williston VT 05495 Www microstrain com inertial 3DM GX3 25
52. st deployments 2 2 1 Record position and orientation of the ADV head For deployments involving cable head IMU ADVs it is critical to record the position and orientation of the ADV head relative to the ADV body pressure case Details of the definitions of these coordinate systems can be found in appendix A The variables should be measured as accurately as possible as errors will propagate through motion correction calculations and lead to errors in the motion corrected velocity measurements As a rule of thumb the vector a should be measured to within a few mm The orientation matrix of the ADV head H should be mea sured to within 2 2 Euler angle degrees 2 2 2 Software configuration At least as important as recording the orientation of the ADV head relative to the ADV body is configuring the ADV to record the correct data channels for performing motion correction Three primary data channels are needed to perform motion correction 1 The linear acceleration vector d is integrated to obtain an estimate of the translational velocity of the ADV body and head 2 The angular rotation rate vector is used to estimate the velocity of the head that is due to rotation of the ADV about the IMU Figure 3 ADVs mounted on a strongback vane prior to deployment The heads and bodies are tilted at 15 to account for mooring blow down NMSS shackles and pear links connect the strongback to the mooring lines The strongback is le
53. tery consumption and memory requirement estimates for the SeMS S SHOWN e endo o a a Be be Be ee oe A eo doe SRM OS eee i Figure 5 An example velocity time series measured using a TTM at Admiralty Inlet A The mean stream wise velocity blue At Smin is over layed on the full signal grey B A 5min data window of the turbulent piece of the streamwise velocity u The dashed lines indicate one standard deviation Figure 6 A time series of turbulence statistics measured from a TTM at Admiralty Head a velocity b turbulence intensity c turbulent kinetic energy and its components d Reynold s stresses Shaded re gions indicate ebb red and flood blue periods where U gt 0 7 Turbulence intensity is only plotted during these periods because it is meaningless for small values of U The mean over the data record is TO ia ee eR ae ee Re Oa SRS a eb goede ee Da ee eS te es Figure 7 A histogram of the mean horizontal velocity magnitude 02202000 Figure 8 A comparison of the shape of spectra at two different sites from ADVs on a rigid tripod A and a TTM B The spectra for each velocity component u v w are in blue green and red respectively The shaded region indicates the inertial subrange in which the spectra decay like f 5 3 and all components have nearly the same amplitude The dashed line indicates a f 5 3 slope The difference in amplitude of the spectra between A and B is expected
54. two basis vectors and phead is into the transducer face For fixed head Nortek Vector ADVs the body frame and head frame have parallel coordinate systems H is the identity matrix and the head frame is translated 21cm along the z axis That is lf a4 0 0 0 21 m Nortek 2005 For cable head ADVs the position and orientation of the ADV head is arbitrary This means that when preparing to make measurements using cable head ADVs the orientation and position of the ADV head must be accurately recorded in order to allow the ADV measurements to be transformed into the body frame during post processing For the example in Figure 12A ae 254 64 165 mm and In general H will not necessarily be symmetric nor will it have so many zero elements i e these characteristics are specific to the head body alignment of the example A 3 2 The IMU coordinate system Like the ADV head the coordinate system in which the IMU measurements are made must be clearly defined and documented In general the IMU frame is related to the body coordinate system by m A a For the Microstrain 3DM GX3 25 IMU as it is integrated into the Nortek Vector Figure 11 ae 0 006 0 006 0 150 m and 0 0 1 A 0 1 0 1 0 0 The DOLfYN software package automatically rotates all IMU vectors so that orientation and motion data returned by dolfyn io read_nortek is in the ADV body frame see the dolfyn io nortek NortekReader sci_microstrain sour
55. vel range m s 5 25 River V IMU AcclAngRMagxF v E Use Advanced Settings Start Update Cancel Apply Help Figure 4 The Nortek Vector program s deployment planning pane with some typi cal settings for quantifying turbulence at tidal energy sites Required settings are highlighted in red and recommended settings are in green The blue box points out the battery consumption and memory requirement estimates for the settings shown B If continuous sampling will deplete available batteries before the end of the deployment use burst sam pling i Setthe Number of samples per burst to capture 10 20min of data ii Setthe Battery pack selection to the batteries that are available iii Adjust the burst interval which must be greater than the sampling period until Battery utilization 1s between 90 and 100 9 Be sure that the memory card in your ADV has sufficient capacity for the deployment See the documentation for details on clearing the memory card if necessary For convenience the DOLfYN software package provides a sample Nortek Vector configuration file Nortek_ Vector_with_IMU dep in the lt DOLfYN root gt config_files ADV folder with the settings described above To use one of these files download it open it with the Nortek Vector software then view and adjust other parameters to fit your deployment needs as necessary The IMU related options will be correctly pre set when usi
56. xed to its head Figure 3 All components of the strongback vane including shackles are constructed from non magnetic materials high density plastics and non magnetic stainless steel so that the IMU compass can accurately resolve the undistorted magnetic field of the earth i e measure North If magnetic materials must be used in the vicinity within 2m of the ADV body the IMU compass should be recalibrated in the presence of those magnetic materials and in the exact orientation as they will be deployed Nortek 2005 At its leading edge flat stock NMSS sandwiches the plastic fin to form the strongback backbone 3 4 holes are drilled through the top and bottom of the backbone so that Non Magnetic Stainless Steel NMSS shackles can connect the strongback to the mooring lines Figures 3 The ADV heads and bodies are tilted 15 from the vertical axis of the strongback to account for the mooring blow down of 10 20 at 2m s 2 2 Instrument configuration When preparing instrumentation for deployment always follow the manufacturers recommendations Be sure to 1 Perform bench tests to confirm the instrument is operating correctly 2 Install batteries with sufficient capacity for the deployment Manufacturers configuration software can aid in determining battery life It is recommended to use non rechargeable batteries because rechargeable batteries can degrade overtime resulting in early instrument shut down and an incomplete
57. y to impart larger forces than shorter ones and longer eddies are likely to be anisotropic such that L gt 6 The first two of these mean shear and Reynold s stresses can be measured using acoustic Doppler profilers ADPs Stacey et al 1999 Thomson et al 2012 The Reynold s stresses turbulence spectra and spatial coherence can be measured using acoustic Doppler velocimeters ADVs This situation requires that detailed turbulence measure ments at tidal energy sites requires measurements using both ADPs and ADVs The deployment of ADPs on the seafloor for this purpose is well described and commonly performed by engineers scientists and ocean professionals around the world While ADVs have the accuracy to measure detailed statistics of turbulence they must be positioned at a point in the flow that is relevant to HKTs i e at hub heights of 10m or more While it is possible to do this using rigid towers such an approach is technically challenging and expensive to implement at tidal energy sites where currents often exceed 3 m s To reduce the cost of hub height turbulence measurements this document details a methodology for making turbulence measurements from moorings This approach is made possible by the recent integration of inertial Turbulent energy and turbulence intensity can be computed from this ule yet Mean Velocity Turbulence lotor Instantaneous Flow Field Mean Flow Profile Figure 1 Diagram
Download Pdf Manuals
Related Search