Wind farm performance validation through machine learning: Sector-wise Honest Brokers

Recent methods to optimize wind farm performance require new methods to assess and validate wind farm level performance. This paper introduces a machine learning approach based on sector wise honest brokers in order to determine expectation of wind energy and validate performance improvements. The approach treats every turbine in the wind farm as a virtual Metmast. Farm level expectation of power is determined based on machine learning models trained on baseline data with input features reflecting “Honest Brokers”: turbines that experience similar conditions in both a training interval and a testing interval in which we are expecting a change in farm performance. Our approach is able to validate farm level improvements even in the face of farm optimization technologies for controlling wakes that change the wind profile within the farm.