Ensuring that all turbines are operating at optimal levels is a goal for all wind farm operators. Poorly performing turbines result in lower revenue, and depending on the cause of the underperformance, they could also increase maintenance costs. Both of these are worth consideration and something to avoid, if possible.
However, detecting underperformance is no simple task and is further complicated as terrain complexity increases, because separating terrain effects on the wind from turbine-performance-related variability becomes increasingly difficult.
A method has been developed that can account for terrain effects on the wind and can then be used to effectively identify turbines that may have performance problems – or, in some cases, outperform the fleet. This methodology has been employed with excellent results at two operating wind farms. The analysis was initially performed in order to normalize terrain effects on the wind – and, thus, turbine power – a crucial step required for wake model validation. In the process of analyzing the observed turbine power at one of these sites, some turbines that were underperforming and others that were performing exceptionally were identified.
An investigation into the possible cause of underperformance came next.
The wind farm that is the subject of this analysis is located in the southern U.S. Great Plains. The site has moderate terrain complexity, yet there are significant performance differences across the site. The project features 66 modern multi-megawatt, variable speed, variable pitch wind turbines.
Of the site’s 66 turbines, 40 are not subject to array effects in the prevailing southerly wind direction. One of the unwaked turbines has had low availability and was not included in this study. The remaining 65 turbine sites were included in the initial analysis, which was conducted for a wake model validation study, but the 39 unwaked turbines are the subject of the analysis presented in this article.
Turbine performance data from the project SCADA system was assembled, and the 10-minute data records were filtered to select only records that met these criteria:
- All 65 turbines had 100% availability,
- All turbines had power greater than 0 kW,
- The wind direction prevailed from the south and
- No curtailment was imposed.
Each turbine’s 10-minute power values were averaged over all available data records that met the filtering criteria. For purposes of anonymity, the observed average power per turbine was normalized by the rated power for this presentation.
Terrain exposure values were calculated, and the average power from the unwaked turbines was analyzed with respect to their respective exposures. Since the observed wind direction for the turbine performance data was southerly, one might expect that the exposure to the south, the upwind direction, would matter the most in terms of affecting the wind that drives turbine power performance.
However, the correlation of the average power vs. upwind exposure had an R2 = 0.124, indicating that the upwind terrain does not affect the wind and, thus, turbine performance to a significant degree. A correlation was also performed between the average power with respect to the turbines’ elevations, and the result was better but still not very good, with an R2 = 0.621. Average wind speeds from nacelle anemometers are used in some instances to gauge turbine performance, but the correlations are not sufficient to clearly show performance anomalies, and an R2 on the order of 0.87 has been calculated under similar conditions at this site.
Analysis of wind data at numerous sites has shown that it is the downwind terrain that is the prime driver in determining how fast the wind blows within a project development area and that the upwind terrain, and perhaps elevation, typically plays a secondary and supporting role. This tends to show up most clearly when the analysis is restricted to a rather narrow directional range, as was done here.
Figure 1 shows the relationship between the average power and downwind exposure for the 39 unwaked turbines. Six of the turbine sites, two groups of three, are separated from the remaining 33, which are used as the control, or “Standard,” group. One group of three is identified as “Over” performers, and the other group of three is identified as “Under” performers, based on the analysis.
In the graph, we observe a very highly correlated relationship between the average power for the 33 turbines in the control group and their downwind exposures, with an R2 = 0.984. This clearly indicates that the downwind terrain has a very high degree of influence on how fast the wind blows and, perforce, how much power the turbines will produce.
However, this relationship is for 33 Standard – or what might be considered as “standard performance” – turbines of the 39 turbines included in the graph. It turns out that the Over group of three are non-standard: One has a larger rotor than other turbines used at the wind farm but the same rated power, and the other two have performance enhancements on the rotor blades. The turbine with the larger rotor has an average power that is clearly well above the trend established by the 33 Standard turbines, and the two with the blade enhancements have average power above the trendline. It is noted that the data set in this analysis includes data records before the performance enhancements were added, so their effect is not indicated as clearly as when the data is filtered again to include only data records when the enhancements were in place.
The average power for the Under group of three turbines falls well below where their exposure, as well as the established relationship between exposure and power performance for the Standard turbines, indicates they should be. Based on this analysis, it is suspected that there are performance problems with these three turbines that have gone undetected. The magnitude of the underperformance is on the order of 5%, based on the performance of the other turbines that have terrain exposures of the same magnitude. Due to the strong relationship between terrain exposure and power output at the Standard turbine sites, there is evidence that the relatively low performance of the Under group is not entirely due to low wind.
An investigation into the possible cause of the apparent underperformance of the three Under turbines was conducted. The concurrent power output data from the three turbines in question, which will be referred to now as Group 1, was compared to the power output from the three adjacent turbines in the same string. These will be referred to as Group 2. The Group 2 turbines are on terrain that has almost exactly the same elevation as Group 1, although the exposures for Group 2 are lower. Based on the observed relationship between turbine performance and exposure, as in Figure 1, the Group 1 turbines should have higher average power than Group 2, but the opposite is the case. The Group 2 turbines are among the 33 Standard turbines.
A third group of turbines, referred to here as Group 3, were also included in this analysis. They are the next three turbines adjacent to Group 2. Thus, Groups 1 through 3 represent nine contiguous turbine sites in the project. The Group 3 turbines are also among the 33 Standard turbines but are in a terrain situation that indicates lower wind, and their performance is consistent with that. The three groups of turbines are marked differently to identify their relative performance in Figure 2.
In this analysis, the average power for the three groups was calculated using a binning process referenced to the average power for the Group 2 turbines, which appear to be operating in a Standard fashion. The concurrent average power from the three groups was calculated in each power bin (bin width of 100 kW, referenced to Group 2) over the range from 100 kW up to full rated power. In each mean power bin, the difference in average power between Group 1 and Group 2 and between Group 3 and Group 2 was calculated. Figure 3 shows the difference in power (▲ power) vs. the average power from Group 2, with some smoothing applied so that the trends are more easily observed. The ▲ power values are expressed as percent of rated power.
It is observed that as Group 2 power output increases up to around 50% of rated power, the ▲ power between Group 1 and Group 2 is quite small, fluctuating around 0%. However, as Group 2 power output increases beyond ~50% of rated power, the ▲ power between Group 1 and Group 2 becomes negative. The observed change in relative performance at ~50% of rated power occurs at around the power level when the blades begin to pitch. This is a clue to a possible cause of underperformance, which is suspected to be related to improper blade pitch control.
However, in the interests of objectivity, one must consider the possibility that, for whatever reason, the observed relationship between average power and exposure exhibited by the 33 Standard turbines may not hold for the apparent underperforming turbines in Group 1 and that what looks like underperformance is really due to lower wind speed. This hypothesis is tested by the inclusion of the Group 3 turbines.
In the case of turbines that appear to be performing properly, but clearly have a less energetic wind resource, the relationship is quite regular and consistent up to the point where Group 2 is producing at approximately 85% of rated power. Above that power level, Group 2 is approaching the knee of the power curve but Group 3 is not, and the relative power difference begins to decrease, as Group 2 approaches rated power.
Considering the results for Groups 1 and 3 relative to Group 2, the behavior of Group 1 is not consistent with low-wind-speed sites and reinforces the suspicion that there are operational problems of some kind that result in power performance that is lower than it should be. Further, more detailed analysis of time series SCADA data has been conducted, which has revealed more clues, but finding conclusive proof of underperformance is not a simple matter, as these turbines are highly complex machines with many interrelated operational components, and they are driven by an inherently chaotic entity – the wind, which adds further complications, particularly over relatively short time periods.
A technique has been developed whereby suspected underperforming, as well as overperforming, turbines can be identified. However, the discovery and proof of the underlying cause of any underperformance, if it exists, are challenging. In the case of the wind farm and the specific turbines that were the subject of this analysis, the original equipment manufacturer was contacted and given the same information as presented here. Their investigation did not reveal anything related to blade pitch that might account for the suspected underperformance. w
Industry At Large: Wind Farm Operations
Using Terrain Modeling To Detect Underperformance
By Jack Kline
Detecting the root causes of wind farm underperformance can be fraught with challenges. However, a new modeling technique may provide some assistance.
NAW_body NAW_body_bi NAW_body_b_i NAW_body_bNAW_body_i