Research of Neural Network Predictive Control in the Variable-Pitch Control System of Wind Turbine

Energy crisis has become increasingly heavier, and the development and use of renewable energy is the primary means to solve the problems. Wind power is a new green energy, with its environmentally friendly, renewable and other characteristics, and it attracts the majority of researchers’attention. Wind generators are the primary means of access to wind energy, with further scientific researches, wind power plants are all over the world. Energy emitted by wind turbines account for the proportion of total global electricity generation is increasing.Variable pitch wind turbine with its good performance and high-speed self-adjusting performance, will gradually replace the fixed pitch wind turbines, and become the main choice of types of wind turbines. Pitch control system for the study has important practical significance. In this paper, a new pitch controller based on short-term forecasts of the wind speed is proposed, in order to avoid the pitch adjustment controller acts too often during operation of variable pitch wind turbine. Firstly, the speed characteristic of the wind is studied, and then a prediction neural network with good performance is build. Secondly, the rules and control process of the neural network controller are designed. Finally, the neural network predictive controller added into the conventional pitch controller to form a new pitch control system.The forecast performance and prediction error are tested by the MATLAB simulation software. The performance of the new built pitch controller is simulated by the PSCAD simulation software. Simulation results show that the method can effectively reduce the blindness of pitch adjustment, reducing unnecessary frequency of the pitch regulation. Meanwhile, this method improves the performance of the pitch controller, and extends the operating conditions of the Wind turbine, and makes its power output more stable.