A Data-Driven Approach for Monitoring Blade Pitch Faults in Wind Turbines

A data-mining-based prediction model is built to monitor the performance of a blade pitch. Two blade pitch faults, blade angle asymmetry, and blade angle implausibility were analyzed to determine the associations between them and the components/subassemblies of the wind turbine. Five data-mining algorithms have been studied to evaluate the quality of the models for prediction of blade faults. The prediction model derived by the genetic programming algorithm resulted in the best accuracy and was selected to perform prediction at different time stamps.