基于AdaBoost与BP神经网络的风速预测研究

Wind Speed Prediction Based on AdaBoost and BP Neural Networks

介绍了基于AdaBoost的多神经网络集成预测方法。集成方法的预测结果优于其他方法的预测结果,这一点在理论上和经验上已经得到证明。AdaBoost是适用于时间序列预测的集成方法。基于AdaBoost算法,采用多个BP神经网络训练随机生成的风速样本,再由多个训练结果生成最终的风速预测值。用该方法预测的误差低于用单一BP神经网络进行的预测,其分析和仿真结果表明了其优越性。

This paper introduces an AdaBoost-based multineural network ensemble method for wind speed prediction. The result of the prediction by the ensemble method is theoretically and empirically proved to be superior to those by other methods. The AdaBoost algorithm is applied to the time series prediction. Based on the AdaBoost algorithm, back-propagation neural networks （BPNN） are generated; each for training on a random set of examples on wind speed data, then the results of each base learner will be combined to form the final hypothesis. The prediction error by this method is smaller than that by single BP neural network, and the analysis and simulation results suggest that the proposed approach results in better performance.