Evaluation of measuring methods for flicker emission from modern wind turbine

Transcription

1 Evaluation of measuring methods for flicker emission from modern wind turbine Leif S. Christensen ), Poul E. Sørensen ), Troels S. Sørensen ), Henny K. Nielsen ) ) DELTA Dansk Elektronik, Lys & Akustik, Erhvervsvej A, DK- Them, Denmark ) Risø National Laboratory, Technical University of Denmark, VEA-, P.O.Box 9, DK- Roskilde, ) DONG Energy, Teglholmen, A.C. Meyers Vænge 9, DK - København SV, Denmark. ) Vattenfall A/S, Støberigade, København SV, Denmark lc@delta.dk ), poul.e.soerensen@risoe.dk ), troso@dongenergy.dk ), hennyk.nielsen@vattenfall.com ) Abstract This paper presents flicker measurements from three of the worlds largest offshore wind farms, where three different wind turbine topologies are used: stall-regulated, fixed speed, doubly fed generator variable speed, and full scale converter. The measuring method described in IEC - [] is evaluated using measurement results from stall-regulated wind turbines. Selecting S kfic /S n is discussed versus the precision of the flickermeter built according to IEC --. The influence of the chosen number of probability bins for evaluation of P st is discussed. Index Terms Flicker, offshore wind farm, flickermeter, precision of measurement. I. INTRODUCTION Modern wind turbines are sophisticated machines where cutting edge technology is used to be able to harvest the most possible energy from the wind. This includes advanced control strategies giving full control over the production at all wind conditions. One of the main issues concerning the power quality of wind energy has always been the variation of the production and resulting flicker emission. This paper presents flicker measurement results from three of the worlds largest offshore wind farms, Nysted Offshore Wind Farm (DK), Horns Rev Offshore Wind Farm (DK) and Burbo Banks Offshore Wind Farm (UK), each of which has different types of wind turbines:. Induction generator, stall-regulated, fixed speed at Nysted Offshore Wind Farm.. Doubly fed induction generator, pitch-regulated, variable speed at Horns Rev Offshore Wind Farm.. Full-scale converter and asynchronous generator, pitch-regulated, variable speed at Burbo Banks Offshore Wind Farm. Based on the results the measurement method given by IEC - [] is evaluated. As expected the old stallregulated fixed speed wind turbine has the highest flicker level, doubly fed wind turbine comes in second and the full converter type wind turbine has the lowest flicker level. There are significant differences in the statistical distributions: For the stall-regulated fixed speed type the flicker is distributed over a wide range, for the doubly-fed generator type the flicker is almost constant, while the full scale converter type has a few extreme values, but generally a low level. All measurements are low and close to the noise floor of the flickermeter instrument, which is build according to IEC -- []. The test specification is given in table, where the flicker severity P st shall be. ±. for each listed case. Calculation of current flicker according to IEC - [] and [] is done by using a fictitious grid. Selecting the short-circuit apparent power of the fictitious grid must result in a flicker level well within the measurement range of the flickermeter [], but not selected too small as this will results in a higher RMS value of the fictitious voltage and a lower flicker coefficient, when normalized with the short-circuit apparent power []. There is a conflict between having a flicker level high enough to get a good result from the flickermeter and the selection of the appropriate short-circuit level of the fictitious grid. The suggested number of classifier levels of the flickermeter is in IEC-,. Edition [] and in IEC-- []. The influence of the number of classifiers will be investigated in this paper. How to use the measured flicker from a single wind turbine to predict the total flicker emission on park level is previously described in [] and []. A. Measuring Campaign II. MEASUREMENT RESULTS Measurements have been made at three of the worlds largest offshore wind farms: Nysted Offshore Wind Farm, Horns Rev Offshore Wind Farm and Burbo Banks Offshore Wind Farm.. ysted Offshore Wind Farm (DK) The offshore wind turbines are placed in a parallelogram consisting of eight rows of nine wind turbines each. The wind turbines were delivered in by the Danish wind turbine manufacturer Siemens Wind Power (previously BONUS). The tower is 9 m high and the rotor has a diameter of m. The wind turbines are fixed speed stall-

2 regulated and have a rated power of. MW. The total output of the offshore wind farm is. MW. Measurements where made in turbine A, A9 and at the connection point for row A at the substation (see Figure ). In this paper only measurements from A are used. C( ) Nysted Offshore Wind Farm A Figure - C( ), Nysted Offshore Wind Farm A Nysted Offshore Wind Farm A C( ) Figure Layout of Nysted Offshore Wind Farm The number of measurements is shown in Figure, measured turbulence intensity in Figure and the measured flicker coefficients in Figure to Figure. Number of measurements Nysted Offshore Wind Farm, A 9 Windbins Figure Wind speed bins, Nysted Offshore Wind Farm A C( ) Figure - C( ), Nysted Offshore Wind Farm A Nysted Offshore Wind Farm A Figure - C( ), Nysted Offshore Wind Farm A Nysted Offshore Wind Farm A Nysted Offshore Wind Farm, A Turbulence intensity [%] % % % % % C( ) % Figure Turbulence intensity, Nysted Offshore Wind Farm A Figure - C( ), Nysted Offshore Wind Farm A The flicker coefficients are correlated to the wind speed, high wind speed results in high flicker coefficients, but not to the turbulence intensity, as shown in Figure.

3 Nysted Offshore Wind Farm A Horns Rev Offshore Wind Farm M C( ) % % % % % % % % % % % Turbulence intensity [%] Number of measurements 9 Wind bins Figure - C( ) as function of turbulence intensity), Nysted Offshore Wind Farm A Figure Wind speed bins, Horns Rev Offshore Wind Farm M The low turbulence intensity does not result in low flicker, as the turbulence intensity decreases with high wind and the flicker coefficients increases. The coefficients are scattered over a wide range for all wind speeds.. Horns Rev Offshore Wind Farm (DK) The offshore wind turbines are placed in a parallelogram consisting of ten rows of eight wind turbines each. The wind turbines were delivered by the Danish wind turbine manufacturer Vestas Wind System A/S in. The tower is 9 m high and the rotor has a diameter of m. The wind turbines are the doubly fed induction generator and variable speed type and have a rated power of MW. The total output of the offshore wind farm is MW. Measurements where made in turbine, and at the connection point for the first two rows at the substation. In this paper measurements only from are used (see Figure 9). Turbulence intensity [%] Horns Rev Offshore Wind Farm M % % % % % % % % Figure Turbulence intensity, Horns Rev Offshore Wind Farm M C( ) Horns Rev Offshore Windfarm M Figure C( ), Horns Rev Offshore Wind Farm M Horns Rev Offshore Windfarm M Figure 9 - Layout of Horns Rev Offshore Wind Farm The number of measurements is shown in Figure, measured turbulence intensity in Figure and the measured flicker coefficients in Figure to Figure C( ) Figure C( ), Horns Rev Offshore Wind Farm M

4 Horns Rev Offshore Windfarm M C( ) Figure C( ), Horns Rev Offshore Wind Farm M Horns Rev Offshore Windfarm M C( ) Figure - Layout of Burbo Banks Offshore Wind Farm Number of measurements is shown in Figure, measured turbulence intensity in Figure and the measured flicker coefficients in Figure 9 to Figure Burbo Banks Offshore Windfarm, BB Figure C( ), Horns Rev Offshore Wind Farm M As it can be seen from Figure to Figure the flicker coefficients are correlated to the wind speed in the range - m/s, whereas the flicker level is almost constant and not correlated to wind speed, when the wind turbine is at full production above approximately m/s. I.e. the correlation depends upon where on the power curve the wind turbine is producing. The turbulence intensity in the measurement is high (>%) which may be caused by the use of the unfiltered nacelle anemometer signal.. Burbo Banks Offshore Wind Farm (UK) The offshore wind turbines are placed in four rows with seven, eight, eight and two turbines. The wind turbines were delivered by Siemens Wind Power in. The tower is. m high and the rotor has a diameter of m. The wind turbines are the full-converter, variable speed type and have a rated power of. MW. The total output of the offshore wind farm is 9 MW. Measurements were made in turbine B, B and at the connection point for this cable at the substation. In this paper only measurements from B are used (see Figure ) Number of measurements 9 9 WindBins Figure Wind speed bins, Burbo Banks Offshore Wind Farm B Turbulence Intensity [%] Burbo Banks Offshore Windfarm, BB % % % % % % Figure Turbulence intensity, Burbo Banks Offshore Wind Farm B C( ) Burbo Banks Offshore Windfarm, BB Figure 9 C( ), Burbo Banks Offshore Wind Farm B

5 C( ) Burbo Banks Offshore Windfarm, BB Figure C( ), Burbo Banks Offshore Wind Farm B Flicker Coefficient Nysted Offshore Windfarm, A Max 99% 9% 9% % C( ) C( ) C( ) C( ) Figure Flicker coefficient percentiles Nysted Offshore Wind Farm A Burbo Banks Offshore Windfarm, BB Horns Rev Offshore Windfarm, M C( ) Flicker Coefficient Max 99% 9% 9% % Figure C( ), Burbo Banks Offshore Wind Farm B C( ) C( ) C( ) C( ) Figure Flicker coefficient percentiles Horns Rev Offshore Wind Farm M C( ) Burbo Banks Offshore Windfarm, BB Figure C( ), Burbo Banks Offshore Wind Farm B The flicker coefficients are generally low but there are a few odd high values in the measurement, which in worst case could result in a high 99 th percentile. The flicker level is not correlated to wind speed and appears almost constant. B. Comparing the three different topologies The three parks use different wind turbine topologies with different flicker characteristics, which will be evident from comparing Figure, Figure and Figure. Flicker Coefficient Burbo Banks Offshore Windfarm, BB C( ) C( ) C( ) C( ) Figure Flicker coefficient percentiles Burbo Banks Offshore Wind Farm B As expected the old stall-regulated fixed speed wind turbine has the highest flicker level (Figure ), doubly fed generator wind turbine (Figure ) comes in second and the full converter type wind turbine (Figure ) has the lowest 99 th percentile flicker level. There are significant differences in the distribution: For the fixed speed type, the flicker is distributed over a wide range, for the doubly fed generator type, the flicker is almost constant, while the full converter type has a few extreme values, but generally a low level. For all measurements the ratio S k,fix / S n =, and for the full converter type some of the measurement are close to the noise floor of the instrument. Max 99% 9% 9% %

6 C( ) III. EVALUATION OF THE METHOD The method for measuring current flicker described in [] is based on a fictitious grid. The fictitious grid is represented by an ideal phase-toneutral voltage source and a resistance R fic in series with an inductance L fic. Flicker is measured on u fic (t), the voltage over the current generator, where the measured current, i m (t), is fed into the simulation. Figure Fictitious grid for simulation of fictitious voltage Figure Simulation of fictitious voltage R fic and L fic shall be selected to obtain the wanted network impedance angle and to represent an appropriate shortcircuit apparent power of the fictitious grid S k,fic. The shortcircuit ratio S k,fix / S n shall be selected so that measured P st values are well within the measuring range of the flickermeter, but not so low that the mean RMS value of u fic will deviate significantly from the mean RMS value of u o. To illustrate this, a number different S k,fix / S n ratios has been used on the same minute data from Nysted Offshore Wind Farm. The resulting flicker coefficients should ideally not be affected by the selected ratio, if the ratio is selected appropriately. The time period is random selected in the data set from Nysted Offshore Wind Farm (see Figure ). P and Q [kw, kvar] - Steady state production Time [s] P [kw] Q [kvar] Wind Speed Figure minute data from Nysted Offshore Wind Farm Only network impedance angle is used in the following analysis, as the angle does not influence the aspects of the method, which is studied. Windspeed [m/s] C Skfic/Sn C( )Ph Figure 9 Resulting C( ) using different values of S k,fic /S n ratio. The flicker coefficient C( ) varies from. at ratio S k,fic / S n = to. at ratio S k,fic / S n =. At ratio S k,fic / S n = the flicker coefficient is.9 and at ratio S k,fic / S n = it is.. The standard [] indicates a range from S k,fic / S n = to S k,fic / S n = and within this range C( ) varies % adding to the inaccuracy of the measurement. The C( ) value is normalized with S k,fic / S n from the measured P st,fic. Figure 9 shows that the relation between P st and S k,fic / S n. is not linear. An appropriate S k,fic / S n. shall be used to assure that the applied flickermeter gives P st,fic values well within the measuring range. To investigate, which ratio is the appropriate to use for the flickermeter, measured P st,fic is plotted as function of S k,fic / S n. Pst,fic Pst,fic ( ) Skfic/Sn Figure Resulting P st,fic( ) using different values of S k,fic /S n ratio. The P st,fic value vary from. at ratio S k,fic / S n = to. at ratio S k,fic / S n =. The test specification for a flickermeter is given in table of [], where the flicker severity P st shall be. ±. for each listed case. The accuracy of the flickermeter is % or better, when testing with a signal giving P st =.. Therefore the flickermeter is in fact not suited for measuring low values and in particular not for values below.. In this case S k,fic / S n > gives P st,fic below.. To justify the use of the flickermeter, it is recommended in [] to use classifier levels instead of, as described in []. To investigate, if this will give a better result, a minute period has been analyzed with to. classifier levels,

7 as shown in Figure, where it is evident, that using more than classifier levels does not improve results. Pst,fic( ) However, with those low flicker values, the flicker emissions from such wind turbines are not likely to cause power quality problems.. Pst,fic Classifier levels Pst( ) Figure Resulting P st( ) using different numbers of classifier levels. The P st,fic varies from.9 using classifier levels to.using. classifier levels. Assuming that the correct value is at classifier levels, the difference is % when using classifier levels and only. % when using classifier levels. The computing time for analyzing minute data increases by % when going from to classifier levels and by % when going from to classifier levels. It is therefore evident that using more than about classifier levels is not justified. IV. CONCLUSION This paper has presented flicker measurement from three of the worlds largest offshore wind farm, where three different wind turbine topologies are used. As expected the old stall-regulated fixed speed wind turbine has the highest flicker level, doubly fed generator wind turbine comes in second and the full converter type wind turbine has the lowest flicker level. There are significant differences in the statistical flicker distribution. For the fixed speed type the flicker is distributed over a wide range, for the doubly-fed type the flicker is lower and at rated power almost constant, while the full converter type has a few extreme values, but generally a low level. Selecting the ratio S k,fic / S n is very important, as it has a great impact on the measured flicker coefficients. Therefore the S k,fic / S n ratio must be stated in the report or set to a fixed value in the standard. If the ratio is unknown, a comparison between different measurements or wind turbines is not possible. Using more classifier levels in the flickermeter does improve the result, but using more than classifier levels does not give a better result. Using the recommended levels instead of increases the computing time % and the resulting accuracy of P st is only improved by.%. Generally the accuracy of flickermeter is not sufficient to measure the very low flicker values from modern wind turbines and better accuracy is needed. ACKNOWLEDGEMENTS This paper describes the measurement system and the measurements performed within the project entitled Voltage conditions and transient phenomena in medium voltage grids of modern wind farms, contract --, funded by Energinet.dk. The project is carried out in cooperation between DELTA (project manager), DONG Energy, Vattenfall and Risø National Laboratory REFERENCES [] IEC -. Measurement and assessment of power quality characteristics of grid connected wind turbines. First edition. Dec.. [] IEC -. Measurement and assessment of power quality characteristics of grid connected wind turbines. Second edition. Aug.. [] IEC --:99/A:, Testing and measurement techniques Flickermeter Functional and design specifications Edition., - [] L. S. Christensen, B. Barahona, P. Sørensen, T. Sørensen, T. Olsen, H. K. ielsen, Power Quality in large offshore wind farm predicting flicker on park level from PQ measurements on single wind turbine, EWEC. [] Braulio Barahona, Power quality of wind farms, validation of standard methods for assessing flicker and harmonics, Master thesis, Risø DTU, June.