Short-term Wind Power Prediction Based On SVM By Improved Genetic Algorithm

Great deal of wind power connected to grid will seriously impact on the safety and stabilization of power grid because of its randomness, intermittence and volatility. Therefore, it is of great significance to improve the wind power short-term prediction accuracy for the real-time scheduling of wind farms and maintenance of the stable operation of the power grid. So far, a lot of research for short-term wind power prediction has been made, but the prediction results have not been ideal.In this thesis, aiming at the shortcomings of the existing methods, a method for short-term wind power prediction based on improved genetic algorithm and optimized SVM is proposed by integrating wavelet theory, genetic algorithm, niche algorithm, immune algorithm and support vector machine algorithm together. The following research work has mainly been carried out:1) In order to reduce wind noise signal of historical data, a smooth processing of the historical data of wind power is made by using wavelet analysis method according to the characteristics of the randomness of wind data sequence.2) The immature convergence problem which occurs during the later period of the operation in the standard genetic algorithm is improved by combining niche algorithm with standard genetic algorithm. And according to the situation where the niche radius value is fixed in the process of operation in the niche genetic algorithm, the niche radius value is made to be updated with the update of population by programming.3) The concept of vaccine is introduced in the niche genetic algorithm, and the initial population is produced by the lead of the vaccine to reduce time and increase optimization speed. And also the scope of the penalty and kernel coefficients and the real-time update of the vaccine are considered with the introduction of the new data samples.4) On the basis of the smooth processing of the historical data and the improved genetic algorithm, SVM short-term wind power prediction model based on the optimization by improved genetic algorithm is established.In order to verify the effectiveness of the proposed prediction model, the model is used to make short-term forecasting of wind speed by the wind speed data samples in a wind farm and that of wind power by a wind power data samples in another wind farm, and the results are compared with those by other prediction methods.The results show that the proposed prediction model has high forecast precision and optimization speed for short-term forecasting of wind speed and power, and the method has good universality and effectiveness.