Research On Prediction And Control Of Wind Power System Based On Improved Neural Network

Wind power is a kind of clean, free and renewable natural resource. Since the 1970 s oil crisis, under the pressure of energy shortage and ecological environment degradation, wind power generation is chosen as a new type of efficient technology has been got the attention of the society from the world. In the future, wind power will also be one of the important methods to solve the problems from energy and environment.In wind power generation systems, the randomness and volatility of wind will affect the quality of power and the reliability of power system. To improve the prediction accuracy of short-term wind speed is of great significance for the real-time scheduling of power, improving the reliability of power supply and lowering the cost of the wind power generation. There has many ways to be put forward to predict wind speed at home and abroad, compared with the merits and demerits, a method based on B-spline neural network optimized by particle swarm optimization is proposed to predict the short-term wind speed. B-spline neural network can change the division of input space and the definition of basis function flexibly, but if the input space can not be chosen properly, BSNN is sasy to fall into local minimum and which will influence the final predict accuracy. Particle swarm optimization algorithm is used to optimazi the nodes of BSNN, which can avoid the BSNN fall into local minimum and improve the prediction accuracy. Simulation results verify the effectiveness of the method.Micro-grid frequency optimization control problem that exists in the wind power and diesel generator hybrid generation system is researched in this paper. Wind power generation adopts the optimal tip speed ratio to track the maximum wind power point. Diesel generator is used to compensate power bias beween loads and wind power supply. The changing of wind speed and loads will make mico-grid frequency have a bias, which will have a bad influence on the generators and electricity system. A frequency controller based on Action Dependent Heuristic Dynamic Programming(ADHDP) is desighed in this paper, RBF neural network is adopted as the function approximation structure of ADHDP. Frequency bias caused by load change with large amplitude and long cycle can be limited in the allowed range by ADHDP controller. Simulation results verfify the effctiveness of the method.In the wind power generation system, maximum wind power tracking method based on optimal tip speed ratio is easy to be realized, but need to measure the wind speed value. To think about this, a method based on cloud RBF neural network and Action Dependent Double Heuristic Dynamic Programming(ADDHDP) is designed to track the maximum wind power point, it does not need to measure the wind speed. Cloud RBF neural network is adopted as the function approximation structure of ADDHDP,which can improve the approximation accuracy and convergency speed. The work principle of ADDHDP is very similar to ADHDP, but ADDHDP is more rigorous in design structure than ADHDP and it has better control accuracy. Permanent magnet synchronous generator(PMSG) is adopted in the wind power generation. Optimal power-speed curve and vector control principle are used to control the electromagnetic torque by adjusting the voltage of stator with ADDHDP controller, so the maximum wind power point can be tracked accurately. Simulation results verfify the effctiveness of the method.