Presentations

Abstract

Operational assessment and monitoring of wind turbine and farm performance is a crucial activity to ensure that wind energy is effective and continues to grow its share in decarbonizing the electricity mix. In that context, determining whether turbine performance meets the specified power curve brings value, to put in place a range of corrective actions, or to adjust financial planning for future years of farm operation. However, with today's operational sensors, it is not straightforward to disentangle the causes of energy shortcomings related to the wind resource, wakes effects and turbine power performance. Indeed, on one hand standard nacelle anemometry is not accurate enough and tends to underpredict the true wind speed. It is also sensitive to the site, terrain, exact turbine configuration, software upgrades and even turbine control actions. On the other hand, pre-construction energy estimates remain too uncertain to be used to draw conclusions for operating wind farms. We propose a new approach to improve this situation, based on the use of physics-based wind and turbine performance models. The principle is to use predicted time series of the wind and power as a reference for each turbine in each site. This enables several applications including power performance assessment and monitoring, which is the focus of this work. Our presentation is divided into three sections. In the first part, we present the underlying principles and methodology to this approach. In the second part, we present validation results against independent operational sensors. In the last part, we present case studies where this approach was applied to operational sites. In the validation section, we use 5 validation sites where we assess the prediction accuracy for the averaged wind speed time series against independent measurements. Having compared the results obtained for different wind flow modeling sources, our results show that the wind speed mean absolute bias can be typically lower than 0.05 m/s, while the correlation coefficient is close to 0.8, which is a solid foundation to enable absolute performance verification. We also report validation results for the joint distribution of wind speed and direction, and for other wind parameters which are important factors for power performance assessment. In the last part, we go over several operational case studies. We first present examples where the approach was implemented to detect under-performance and draw similar conclusions as when installing a third-party sensor. Finally, we present a case study where an incorrectly implemented curtailment scheme was detected. In conclusion, we present a broader list of applications which can benefit from this solution.