A New Condition Monitoring Method for Wind Turbines Based on Power Curve Model

Abstract

A new data-driven condition monitoring method for wind turbines is proposed to prevent a turbine failure in a wind farm. The method works through the power curve model which is built with historical SCADA data. New constraints that established based on Betz' law and RC model are developed for governing the power curve model. Since abnormal data has a strong impact on the power curve model, the Inner-DBSCAN algorithm is proposed to reject it. In addition, we use an edge recognition method for normal data to form the power curve model. Then, the turbine operation condition can be monitored through the model. Its effectiveness through industrial studies is confirmed, and the time cost of building the power curve model is only 1.08s. © 2016 SERSC.