Health Monitoring for Variable Pitch Systems of Wind Turbine Using Multi-layer Perceptron Strategy

This paper combines the artificial intelligence algorithm with the modeling and simulation for variable pitch systems of wind turbine, and proposes a fault prediction using the artificial neural network technique of multi-layer perceptron. The operating principle of the variable pitch system is analyzed by building the dynamic model, and the fault detection data of the system is obtained. Then, the artificial neural network of multi-layer perceptron is trained with limited data set. The estimated value of pitch angle can be accurately predicted. The residual error can be obtained by comparing the pitch angle predicted by the artificial neural network and the actual pitch angle output of the system model. The system fault detection results can be obtained according to the fault decision-making index. The proposed method can detect and isolate the fault of complex pitch system accurately with uncertain factors, such as disturbance, and predict the change of fault amplitude effectively.