Identification of the wind turbine system based on RBF neural network
YANG Zhen yu1, WANG Qing1,2, WEI Xin gang1, YING You1,2, SUN Yong1,2
(1.Zhejiang Windey Co., Ltd., Hangzhou 310012, China; 2.State Key Laboratory of Wind Power System, 
Hangzhou 310012, China)


Abstract: Aiming at the problems of difficult to establish the accurate mathematical model of wind power generation, identification of the wind turbine based on RBF neural network was presented. The dynamic process of the torque loop and the pitch loop was simulated, RBF neural network algorithm was adopted to identification the torque loop and the pitch loop. RBF basis function was adopted to form space. If the hidden layer RBF parameters was determined, the nonlinear mapping relation was determined. The output layer was the hidden layer nodes output linear weighted summation. The result indicate that identification of torque loop, the input is torque, the output is speed, the torque loop error rate is about 1%. Identification of pitch loop, the input is pitch angle, the output is speed, the pitch loop error rate is about 3%. The pitch loop is a very complicated nonlinear model, the model structure is influenced by many aspects, identification result error is bigger than the torque loop identification error, but the error rate is allowed. The algorithm has higher precision and efficiency.

Key words: the wind turbine; radicalbasis function(RBF) neural network; identification