Wind Turbine Generator Winding Temperature Prediction Based on XGBoost and LSTM

Abstract

Generator stator winding temperature is a significant representation of the health status of wind turbines. Accurate prediction of winding overheating can help us timely formulate operation and maintenance plan and find out the fault source. A combined model is proposed to predict the stator winding temperature of wind turbines based on the weighted fusion of XGBoost (eXtreme Gradient Boosting) and LSTM (Long Short-Term Memory), and the difference in model structure between the two methods is used to improve the accuracy of the fusion prediction results. The SCADA data from on-site wind farm verifies that the proposed combined model can effectively predict the winding overheating, which is of great use for further engineering application.

Keywords

• wind turbine
• winding temperature prediction
• xgboost
• lstm