Use case: Root-cause change of Performance

Turbine model: Vestas V90-3.0 MW

We at ExpertWind quantify Performance by combining machine learning and AI with wind turbine modelling to address most of the challenges associated with SCADA data—ensuring quality, stability, and correcting wind speed measurements when necessary. The result is a reliable, scalable Performance monitoring able to separate real from apparent underperformance on wind turbines.

Fig. 1 - Performance monitoring based on SCADA (blue) and ExpertWind (green). Results were obtained from ExpertWind’s dashboard.

In this example, an operating V90-3.0MW presented a performance that appeared to be deteriorating over time. This behaviour was seen from a significant shift of the SCADA power curve visible on the Normalised SCADA PI showed in Fig. 1.

Fig. 2 - Average wind speed correction time series. An increasing underestimation of the wind speed was detected until it reached 0.7 m/s. After the wind sensor was replaced, the correction parameter went back to normal and close to 0.0 m/s

Our monitoring prediction models detected a change on the wind speed measurements starting around the same date and with a similar trend as the turbine’s Performance Index (PI). The wind speed correction factor (Fig. 2) continuously increased until reaching -0.7 m/s. This behaviour is coherent with a faulty wind sensor that is underestimating the wind speed.

By correcting the SCADA wind speed, we obtain the ExpertWind Performance Index (Fig. 1) indicating that the wide majority of the Power Curve change was actually caused by the faulty wind sensor and not by a real underperformance of the turbine.

A residual underperformance (3-5%) remained after the wind speed values were corrected. This was found out to be related with a pitching issue on the turbine created indirectly by the underestimation of the wind speed.

Once the issue was detected and communicated to the OEM, the wind sensor was replaced (Jan. 2024). The turbine is now measuring the wind speed correctly, the power curve went back to normal and the real underperformance (between 3% and 5% of AEP) was corrected and is in no longer impacting the turbine.

Want to learn more about our Performance monitoring? Feel free to read our article that deep-dives into our method and added value or Contact Us for a demo of our solutions.