A Context-Aware Method For Estimating The Performance Decline With Age Of A Fleet Of Wind Turbines

Wind turbines, like any technical apparatus, undergo performance decline with age but, despite such technology is mature, there are no physical models quantifying the expected magnitude of such phenomenon. Hence, the only possibility is learning from data, which has become possible due to the fact that a vast number of wind farms worldwide has been reaching the end of expected lifetime. Based on such considerations, in the present work, a method for quantifying the performance decline with age of wind turbines is proposed. The qualifying point, which enhances the state of the art, is that the proposed approach is fleet-wide and context-aware, in order to take into account that sensors accuracy affects both the efficiency of wind turbines and the evaluation pipeline of the efficiency. Hence, as a preliminary step, a fleet-wide control of anemometer losses of calibration is set up and, in case any is diagnosed, the corresponding wind speed measurements are consequently corrected. Upon this, a multivariate regression model is set up for each wind turbine, having multiple environmental variables as input and the power as output. By simulating the output for all the data sets of interest and elaborating on the residuals between measurements and model estimates, the average yearly performance decline with age can be estimated. A test case featuring 11 years of data of 11 Vestas V52 wind turbines is discussed. The main novel conclusion is that the wind turbines showing a higher performance decline with age are those who have undergone anemometer loss of calibration. This sheds light on the fact that static errors affecting wind turbine control are an important cause of producible energy loss.