This paper presents the analysis of the study of the flicker emitted from the 1.2 MW tidal energy converter (TEC), SeaGen, against varying tidal parameters. This paper outlines the main elements of the TEC itself, the environment it is located in and the measurement set up. In this paper, the flicker emitted by the TEC is compared with the different tidal parameters, including flood and ebb tides, tidal speed, water depth and turbulence strength and intensity. Flicker emissions have been calculated from measured data in over 90 measurement (10 min) periods, and all of the tidal parameters vary significantly over that testing period. This allows for a detailed statistical and graphical analysis of the variation of flicker with the variation of the tidal parameters outlined above. It is found, with the exception of tidal speed, that there is no strong relationship between flicker emissions and any other tidal parameter. As SeaGen is an asymmetrical TEC with full blade pitching for flood and ebb generation, it was also found that the expected difference of flicker emissions owing to the effect of the submersed crossbeam was not significant. The TEC harmonic performance versus tidal speed is also presented.
SeaGen is a tidal energy converter (TEC) under development by Marine Current Turbines (MCT) Ltd. Previously, MCT sea tested a 300 kW prototype that was installed in Lynmouth in the UK, and which operated between 2003 and 2006. SeaGen marks MCT’s move up to a full-scale TEC technology demonstrator with a rated output of 1.2 MW. SeaGen is an in-stream TEC comprised two horizontal axis turbines with full blade pitch control.
Between August and November 2009, a project was undertaken by ESB International, Chalmers University and MCT to measure, analyse and characterize the power-quality performance of the SeaGen device, with a particular emphasis on developing an understanding of the levels of flicker emissions during operations. This paper summarizes and presents the results from those activities.
2. Environment and technology
SeaGen was installed in Strangford Lough, Northern Ireland, UK in the summer of 2008. It was then commissioned for service over the following months, and has been in regular production since May 2009.
(a) Strangford Lough
Strangford Lough is a 150 km2 sea lough located in the northeast corner of the island of Ireland. The only entrance to the lough is a 7 km channel. Approximately 3 km from the open sea, the channel narrows to about 500 m for 2 km (called ‘The Narrows’) and it is here that SeaGen is located (figure 1).
The tidal current in this channel is one of the fastest currents around either Ireland or Great Britain, with over 350 million m3 of water entering or leaving the lough in each tidal half cycle. The flow peaks at over 4 m s−1 (7.8 kts) in springs. Tidal flows are semidiurnal with two ebb and flood tides per day.
The water depth at the SeaGen location at the lowest astronomical tide is approximately 24 m. Mean low water springs are +0.5 m and mean high water springs are +4.1 m.
Previous tidal current surveys of ‘The Narrows’ indicated that the site chosen for SeaGen would not have unduly turbulent flows because of sea-bed or shoreline features. As the TEC is located nearly 4 km from the open sea, wave action is normally very limited and confined to high-frequency locally generated small waves.
The general structure and outline of SeaGen is shown in figure 2. It has a central column that can be attached to the sea bed either by means of a monopile or, as in the case of Strangford, a quadropod foundation. Supported on the central column is a crossbeam of 27 m that supports the two horizontal axis turbines, each with a swept area of 16 m diameter. The turbines have two blades that can pitch through 270°, with blade pitching replacing yaw control for flood and ebb production. The turbines are coupled to a planetary gearbox that increases the shaft speed to a nominal value of 1000 r.p.m. at the two asynchronous two-pole 600 kW generators. All of this equipment is contained in submersed drivetrains at the ends of the crossbeam. There is a lifting mechanism on board the TEC to lift the crossbeam and drivetrains above sea level for maintenance activities.
The output from the two generators is fed to two full back-to-back insulated gate bipolar transistor (IGBT) converters, and their grid synchronized outputs are ganged together before the step-up transformer. All of this equipment is located within the central column. Power is exported then by means of a submarine cable to the local 11 kV network. The outline of the power take-off (PTO) system is shown in figure 3.
The turbines operate at variable speed below rated tidal current according to the determined optimum Cpλ curve (the ratio of blade tip speed to tidal speed that produces the highest blade efficiency). They operate at a nominal fixed speed above the rated tidal current. SeaGen’s cut-in and rated tidal current speeds are approximately 0.8 and 2.5 m s−1, respectively.
(c) Medium voltage grid connection
SeaGen is connected by way of a submarine three-phase cable to the local 11 kV grid. The grid connection point is fed by a three-phase 11 kV overhead line from a 33 kV subtransmission substation. The source impedance, provided by the Grid operator, at the grid connection point, or point of common coupling (PCC), is 6.28+j5.05 Ω, giving a short-circuit level of 15 MVA with a grid impedance angle of 39°. The source impedance at the terminals of SeaGen (the measurement point 1 (figure 3)) was 7.29+j8.62 Ω when the submarine cable and the 0.69/11 kV transformer are included. This gives a short-circuit level at the measurement point of 10.7 MVA with a grid angle of 50°.
Apart from the measurements taken on board the TEC at the LV side of the transformer, an additional PQ analyser was installed at the grid connection point onshore (see figure 3, measurement point 2), and it carried out a full measurement of the PQ aspects at that point fully complying with IEC61000-4-30 class A . Grid connection voltage remained within 11 kV (+7%, −1.8%) at all times whether SeaGen was operating or not, and system frequency always remained within 50 Hz (+0.5%, −0.8%). When the device was not operating or switching, the typical background Pst and Plt values at the grid connection point were both approximately 0.2. However, they did on a few occasions peak up as high as 4.5 and 2.0, respectively, when SeaGen was neither operating nor switching. The reasons for these spikes in background flicker levels were not further explored, but as they were generally single recorded points, they may have been erroneous recordings. Full details of this analysis is provided in the study of MacEnri et al. .
3. Test equipment
The electrical measurement points for the characterization of SeaGen’s PQ parameters are indicated in figure 3, measurement point 1. The PQ analyser installed at the PCC (measurement point 2) characterized the network’s PQ parameters when SeaGen was not operating, as well as the PQ impact of SeaGen at that location when it was operating. This served as a useful cross check against the calculated, and therefore expected, impact of SeaGen on the local medium voltage (MV) grid.
(a) Voltage and current at measurement point 1
The three-phase instantaneous voltage and current waveforms were recorded using a Unilyzer 902 PQ Analyser set up to record on six channels simultaneously at a sampling rate of 6.4 kHz. The instrument complies with the relevant the International Electrotechnical Commission (IEC) standards [1,3]. Each channel was fitted with a suitable anti-aliasing filter. The specified accuracy of the recorder is ±0.1%.
The voltage was measured through a VT with a ratio of 690∖110 and an accuracy class of 0.5, and was fully compliant with IEC60044-2 .
The current was measured using Rogowski coils with a working range of 10–2000 A (10 Hz–15 kHz) and a specified accuracy of ±0.5% of reading, ±0.5 A. The phase error introduced by the current transducer is less than 1° up to 2.5 kHz and complied with IEC60688 .
(b) PQ analysis at measurement point 2
A Unilyzer 902 PQ Analyser was employed at the PCC. In this instance, it was set up to record the PQ parameters as per IEC6100-4-30  and IEC6100-4-7 . The three-phase voltages and currents were sampled at 6.4 kHz, but only 10 min average values were recorded. In the study of MacEnri et al. , the complete output files (covering periods of more than 4 weeks each) were fully analysed and reported against the limits allowed under EN50160(MV) . The TEC was fully compliant with this standard.
The voltage was measured through a 6350∖110 ratio VT with an accuracy class of 0.5, and was fully compliant with IEC60044-2 .
The current was measured using Rogowski coils with a working range of 2.5–500 A (10 Hz–15 kHz) and a specified accuracy of ±0.5% of reading, ±0.5 A. The phase error introduced by the current transducer is less than 1° up to 2.5 kHz.
(c) Tidal current speed measurement
The tidal current speed was measured using the Valeport Model 803 electromagnetic current meter. The raw data are sampled at 96 Hz and averaged with the digital output transmitted to the control system and historian at approximately 1–2 Hz. Its range is ±5 ms−1, resolution 0.001 m s−1 and accuracy ±1%.
As the current meter is located on the crossbeam and subjected to some accelerated flow, a calibration exercise was undertaken previously against an acoustic Doppler current profiler (ADCP)1 located nearby, and a suitable offset is applied for each velocity level. The current meter is primarily used in the TEC’s operation, protection and control systems.
Voltage flicker is almost and entirely a human perception or an irritation problem. A cyclically varying voltage source supplying the power to incandescent bulbs will cause a cyclical variation in the bulbs’ luminescence that, above a certain level, will be perceived by many people and may cause irritation. This generally results in complaints to the electric utility operators to alleviate or remove the problem. Scientific studies have shown that the level of variation of luminescence that people will tolerate is a function of the frequency it is occurring at. This is defined in the flicker curve (figure 4), and it is shown from 1 mHz up to 15 Hz and varies from 3 per cent voltage variation at 1 mHz and a minimum of 0.3 per cent variation at 8.8 Hz (a particularly sensitive frequency as it coincides with the brain’s alpha waves as it is associated with inducing hypnosis in some people subjected to it).
(b) Flicker metrics
The perception by people of variation in an incandescent bulb’s luminescence is a complex process that involves the bulb, eye and brain, and while being a function of frequency, it is also a function of the duration of its occurrence. In order to standardize the measurement of flicker levels that took account of all of these processes, a standardized flickermeter was developed and is defined under the IEC flickermeter standard . As the voltage variations are effectively amplitude modulated on top of the stable 50 Hz voltage waveform, the instantaneous voltage is input to the meter, demodulated to produce the varying element and weighted according to the eye–brain response. This outputs the instantaneous flicker level, which is then statistically summarized over 10 min giving, ultimately, a single value, called Pst, that characterizes the level of flicker being measured over that 10 min period. Pst is the short-term flicker level over a 10 min period and, owing to the weightings applied in the statistical summary, a value of 1 means that it is at the level defined in the flicker curve in figure 4. A long-term 2 h metric of flicker levels is also widely used for long processes, called Plt, and it is essentially the cubed root of the average of the cubed Pst values in that 2 h period. This flickermeter standard is fully implemented in most power-quality meters today.
(c) Levels of flicker
Utilities are very familiar with the issue of flicker, as in the past, flicker was normally a problem caused by varying loads in industrial processes that created voltage changes in the local surrounding network. Arc furnaces and welders were typical causes of significant flicker problems with Pst values often approximately 1.5 to 2.5. Values greater than 1 will normally cause utility customers to complain to the utility and require action to be taken to remove or mitigate the source of flicker. Typically, in an electric grid today, there would be background Pst and Plt flicker levels of between 0.2 and 0.4. Traditional generators such as hydro, gas and steam plants caused almost no flicker as they had very stable outputs. With the advent of wind-power generators (and other renewable sources such as wave and tidal), they are now also becoming a source of flicker. This is because wind turbines have a variable output, which is because of the natural variation in wind input, wind turbulence and the wind turbine’s uneven power output as it completes each rotation. In grid code requirements for generators or loads connecting to the HV system, typically they would require them to have a Pst value of less than 0.8 and a Plt value of less than 0.6, and utilities will assess such generators and loads to ensure that they will not cause flicker problems on their network.
(d) Flicker coefficients
The level of flicker produced by a load or generator is linearly related to the short-circuit capacity (with its Thévenin equivalent impedance) of the grid to which it is connected and also to the ratio of resistance to inductive reactance (X/R ratio or grid angle) in that equivalent impedance. In order to have a useful metric that can be used to predict the flicker levels caused by wind turbines at any grid impedance or grid angle, the flicker coefficient, Cf, is used. The Pst values created by the load or generator are measured for the standard grid angles of 30°, 50°, 70°and 85° by using the instantaneous measured values of current that are fed through a fictitious grid with no flicker and a corresponding impedance and then normalized (multiplied by the ratio of the short-circuit capacity of the grid to rated capacity of the load or generator). To calculate the actual predicted flicker values on a different grid connection point, the flicker coefficient is calculated from these four values at the new grid angle (linearly interpolated) and is divided by the actual ratio of the short-circuit capacity of this grid connection point to the rated capacity of the load or generator to be installed.
(e) Wind turbine flicker
Early wind turbines were generally fixed speed, stall regulated with a grid-connected induction generator. Their power-quality performance was poor, with output power fluctuations up to 20 per cent of average power reported . Depending on the X/R ratio of the grid connection point and the short-circuit ratio, the flicker levels (Pst) these fluctuations would produce are, typically, in the range of 1–2, which would be well above the acceptable level to a electric utility. For this and other PQ reasons, wind turbines had a poor reputation among utility engineers. Modern wind turbine designs commonly have pitch regulation, variable speed with a doubly fed induction generator or full converters, and these have much lower power fluctuations  and much better all round PQ performance. Typically, they produce Pst flicker levels of around 0.2–0.4 during continuous operations.
5. Assessment methodology
As no standard exists for the measurement and assessment of power-quality performance of TECs, the authors chose to use the closest applicable standard and modify it where necessary. To this end, the existing standard used to measure and assess the power-quality characteristics of grid-connected wind turbines IEC61400-21  was chosen.
The active and reactive power, voltage fluctuations and harmonics PQ characteristic parameters were assessed in this study following the methodology in IEC61400-21 , except where deviations were required, and these are specified below. Ramp rates, set-point control, grid protection and reconnection times were not considered as part of this study. Full details of the methodology are provided in the study of Thiringer et al. .
Voltages and currents on board the device were recorded in July, August and November 2009 at the location indicated in figure 2. In total, there are 89 recordings (10 min long) used in the analysis. These vary in tidal current speeds from around 0.9 up to 3.8 m s−1. In 49 of the data files, both turbines were operating and in the remaining 40, only one turbine was operating. It was not specifically intended to carry out so many recordings with only one turbine operating but, given the methodology employed in IEC61400-21 , results are equally applicable with either one or both operating. It also allowed for comparison of flicker performance between the two operating modes.
(a) Tidal speed measurements
Within each 10 min recording used in the analysis, there were typically between 200 and 600 recorded measurements of tidal velocity. The number varies as the tidal speed is not recorded by the SeaGen system historian at regular intervals, but only at points where the value changed significantly on the previously held value.
IEC61400-21  requires the wind speed to be measured at hub height and recorded at intervals of at least 1 Hz and reported as the average value for the relevant reporting period (10 min in the case of flicker coefficients). A very similar methodology has been applied here given that the current meter is located at hub height and it is updating between 0.34 and 1 Hz. Furthermore, in IEC61400-12-1 , there is an additional requirement that the anemometer response to vertical wind components attenuates in an approximately cosine fashion from the horizontal components. The current meter, which is measured along the principal TEC horizontal axis (and normal to it), has a better attenuation of vertical components as a consequence of its different physical measurement method, which results in an attenuation that is very close to a cosine.
The raw data from the current meter are not filtered. The arithmetic mean value is the representative value for the tidal current speed in the 10 min measurement period following the methodology in IEC61400-21 . The number of tidal current measurement points in the 89 measurement files ranges from approximately 200 to over 600 points (this varies because of the non-fixed recording intervals). The standard error of this mean value as an estimator of the true tidal speed mean in the measurement period is very small owing to the large number of measurement points, and it is less than 1.5 per cent in all measurement files.
Measurement at the single point of hub height is the standard industry practice in the wind industry [10,12] as the representative value for the entire surface area of the rotor disc, and the error this introduces because of the wind gradient is small. The error because of the velocity gradient in water in comparison is even less pronounced, as the much smaller rotor disc diameter has a significantly greater impact in reducing the error than any small increase in the error because of the different velocity gradient in the water compared with air.
It is a requirement in the wind standards that the wind measurement not be obstructed or unduly influenced by structures such as meteorological towers or wind turbines. In the case of SeaGen, the current meter does experience accelerated flows because of the obstruction caused by the cross arm. A calibration exercise was previously conducted using an ADCP located nearby that was not influenced by the TEC itself at various tidal speeds. The values recorded are the corrected values allowing for any acceleration because of the cross arm.
The tidal current measurements in figure 5 show a slight increasing trend. This is because of the accelerating flow in the tidal cycle in this example. Other measurement files show similar trends, with the trend increasing in some cases and decreasing in others. This is a real change in the actual measurement conditions but, given the requirement for 10 min measurement periods for flicker calculation and the nature of tidal flows themselves in Strangford Lough, which have very fast acceleration, it means that such underlying changes in the measurement conditions are inevitable. The average change in mean velocity (based on the linear trend lines such as those shown) was normally less than 3.5 per cent of mean speed, but was over 10 per cent in some of the spring tide measurement files. Such underlying trend increases in mean wind speeds would also inevitably happen during 10 min measurement periods, but there is no requirement in the existing standards of the need to adjust results to allow for this. The same approach has been adopted in this study.
The authors were interested in further analysing the relationship between various recorded tidal parameters and the flicker levels recorded. The results of these studies are presented below.
(b) Active power
The P600 output power from SeaGen is shown in figure 6.
At rated output, the reactive power was less than 60 kVAr, giving a power factor of 0.99.
Thiringer et al.  have outlined in detail how the power-quality standard for wind turbines  was adapted to establish the power quality and, in particular, the flicker performance of SeaGen. It explains how the flicker coefficients were calculated from measurements of voltage and current during continuous operations.
(c) Mean speed
The results of mean tidal speed versus flicker coefficients at measurement point 1 are shown in figure 7 for a grid impedance angle of 50° (the actual angle).
The results show a strong determination between mean water speed and flicker emission of over 94 per cent for the best-fit power trend line shown. The flicker values show significantly higher levels of scatter above the rated velocity of 2.5 m s−1. This is probably due to the effect of the different control strategy as the blades are pitched to shed power. The best-fit trend line is almost an exact squared power relationship between mean speed and flicker coefficients. This latter relationship is also shown by the squares in figure 7.
(d) Turbulence strength
Turbulence strength (Ur.m.s.) is frequently used in fluid dynamics as a measure of fluid turbulence and is defined as
This metric is basically the standard deviation of the measured values about the mean of the sample. Using this metric, the turbulence strength effect on flicker is displayed in figure 8.
Both the linear regression and power curve (shown) have a coefficient of determination between turbulence strength and flicker coefficient of over 55 per cent. From the scatter of values on the graph, it is clear that there are significant factors other than turbulence strength alone contributing to flicker. Thiringer et al.  showed the frequency spectra of the output power and the impact of the crossbeam is clearly visible on the output at a fundamental frequency of 0.47 Hz. The harmonics of this fundamental up to the ninth are also clearly visible. As these frequencies are occurring within the frequency range of the defined flicker curve, they are also contributing to the level of flicker at the output of the TEC, but would not be reflected in the Ur.m.s. values. The different control strategies above and below rated tidal speed may also account for some of the flicker, and this would also not be reflected in the turbulence strength as measured.
(e) Turbulence intensity
The principal metric used in IEC61400-21  for wind turbulence is turbulence intensity (TI) defined as
For wind, at a mean speed range of around 1 m s−1, the TI is initially high at around 50+%, but typically drops dramatically by half or two-thirds by 5 m s−1. The values tend to be very dispersed and scattered in this speed range. The TI drops further at higher mean speeds and levels off at mean speeds of around 12 m s−1 and higher. It is this stabilization of the value that makes it a useful metric for comparing the turbulence of one site with another. In the case of tidal flows, the mean speed range is much more limited, being between 1 and 4 m s−1 typically. Using the metric of TI, the level of flicker at increasing TI values is displayed in figure 9.
Two things can be noted from figure 9. Firstly, just as in low-speed wind, the values are very scattered and, secondly, they are continuing to decrease and have not stabilized within the speed range. This gives the impression that the flicker levels reduce as the TI increases. However, this anomaly is explained by the fact that TI is inversely proportional to the tidal speed, as shown in figure 10 (as it is for wind in the lower mean speed ranges).
The negative correlation between mean speed and TI is because of the fact that the mean speed increase is greater than the corresponding increase in the turbulence strength (Ur.m.s.). This is shown in figure 11.
Figure 11 demonstrates that for a 100 per cent increase in mean speed, the corresponding increase in Ur.m.s. is only approximately 55 per cent on average.
(f) Flood versus ebb tides
Because of the different nature of the flows between flood tides and ebb tides, and the asymmetric location of the crossbeam in relation to the turbines for flood and ebb tides, it was considered probable that there would be a significant difference between flicker levels for the two tides.
As shown in figure 12, this analysis indicated, surprisingly, that there is no significant distinction between flood and ebb tide flicker levels. The only significant distinction between them is that tidal currents achieve higher speeds with flood tides so the highest flicker levels are higher, but this is as expected.
(g) Tidal height
The water height varies from approximately 24 up to 28 m. Figure 13 displays the relationship between water height and flicker levels.
It can be noted that the lower flicker levels tend to occur at the lowest and highest water heights. These would correspond with periods close to low water and high water for the tides, and hence velocities would be lower, consequently giving lower flicker levels. Apart from this, there is no significant or obvious correlation between tidal height and flicker.
As SeaGen has full back-to-back converters, the harmonic performance of the TEC is determined by the harmonic performance of the converters themselves. The voltage and current total harmonic distorted (THD) output of the TEC are shown in figures 14 and 15.
The THD of voltage never rises above 1.25 per cent for any tidal speed and the current THD never rises above 0.65 per cent (two-turbine operation) for any speed.
Overall, SeaGen, from a power-quality perspective, performs very well, and the levels of flicker emission from it are low. The following detailed conclusions are also in order:
— the absolute level of flicker emission is closely correlated with the mean tidal speed squared;
— turbulence strength increases with increasing mean tidal speed, and it appears to be the principal cause of the increasing flicker levels;
— flicker emissions appear to be fairly closely correlated with the near square of the turbulence strength;
— with a very low mean tidal speed range compared with wind, the TI is very scattered and decreases with increasing mean tidal speed and is, therefore, not a useful metric of turbulent flow for TECs;
— turbulence strength is the preferred metric of turbulence for TEC power-quality performance assessment;
— given the asymmetric arrangement of SeaGen’s turbines with respect to ebb and flood tides, there does not appear to be any significant difference in flicker emissions as a consequence of this;
— tidal height has no significant correlation with flicker emissions; and
— THD of voltages and currents are well within acceptable standards for all mean tidal speeds.
The results obtained in this research work indicate that, as tidal velocity has the only significant influence on flicker levels, future sites with lower maximum tidal speeds will probably have even better flicker performance than in Strangford Lough in this respect. However, this needs to be weighed against the fact that future sites will often be in open waters, with significant wave regimes, and these long period waves are very likely to have a significant adverse impact on flicker levels. The exact impact of that will not be known until a SeaGen TEC is located in such a wave regime. However, these results are likely to be readily applicable to all sites that have similar low-wave regimes to Strangford’s.
We thank the management and staff of MCT for allowing access to SeaGen to carry out this study and for all support received during it. The support of ESB International for this work is greatly appreciated and, in particular, we thank James Gallagher, John Whelan and John Kelleher of the Power Systems Studies Team for carrying out the field measurements and some of the subsequent data processing.
1 This is a well-established and accurate water current measurement system that sends out acoustic signals (pings) and calculates tidal flows at different levels by determining the Doppler shift in the sound reflections off microscopic particles moving in the water.
- 1IEC6100-4-30.2008Electromagnetic compatibility (EMC) part 4–30: testing and measurement techniques—power quality measurement methods. Google Scholar
- 3IEC6100-4-15.2010Electromagnetic compatibility (EMC) part 4–15: testing and measurement techniques—flickermeter functional and design specifications. Google Scholar
- 4IEC60044-2.2003Instrument transformers—part 2: inductive voltage transformers. Google Scholar
- 5IEC60688.2002Electrical measuring transducers for converting a.c. electrical quantities to analogue or digital signals. Google Scholar
- 6IEC6100-4-7.2003Electromagnetic compatibility (EMC) part 4–7: testing and measurement techniques—general guide on harmonics and inter-harmonics measurements and instrumentation, for power supply systems and equipment connected thereto. Google Scholar
- 7EN50160.2010Voltage characteristics of electricity supplied by public distribution systems. Google Scholar
Gerdes G& Santjer F. 1994Power quality of wind turbines and their interaction with the grid. Proc. of the European Wind Energy Conf. (EWEC ’94), Thessaloniki, Greece, October 1994, pp. 1112–1115. Google Scholar
Larsson A. 1996Flicker and slow voltage variations from wind turbines. Int. Conf. on Harmonics and Quality of Power (ICHQP ‘96), Las Vegas, NV, 16–18 October 1996, pp. 270–275. Google Scholar
- 10IEC61400-21.2008Wind turbines part 21: measurement and assessment of power quality characteristics of grid connected wind turbines. Google Scholar
Thiringer T, MacEnri J& Reed M. 2011Flicker evaluation of the SeaGen tidal power plant. IEEE Trans. on Sustainable Energy, October 2011, vol. 2, no. 4. Google Scholar
- 12IEC61400-12-1.2005Wind turbines—part 12–1: power performance measurements of electricity producing wind turbines. Google Scholar