提出一种优化的支持向量机风速组合预测模型,首先通过模糊层次分析法对参与组合的单项预测模型进行遴选,在当前风速样本集下自适应决策预测效果较优的单项预测模型的输出值作为支持向量机的输入,将实际风电场风速值作为支持向量机的输出,并采用粒子群算法优化支持向量机组合模型的参数。基于实际运营的风电场数据进行仿真分析,自适应遴选出BP神经网络、RBF神经网络、小波神经网络和遗传算法优化BP神经网络这4种单项预测模型参与支持向量机组合,结果表明所提方法的预测精度不仅高于单项模型,且高于线性组合预测模型和神经网络组合预测模型。 An optimal combination forecasting model of wind speed of support vector machine is proposed. Firstly, fuzzy single analytic hierarchy process (AHP) is used to select the single forecasting model which is involved in the combination. Under the current wind speed sample set, the output of single forecasting model with better predictive effect is selected as Support Vector Machine (SVM) input, wind speed value of actual wind farm is used as the output of SVM, and Particle Swarm Optimization (PSO) is used to optimize the parameters of support vector machine (SVM) combined model. Based on the actual operation of the wind farm data simulation analysis, adaptive selection of BP neural network, RBF neural network, wavelet neural network and genetic algorithm to optimize BP neural network which four single prediction model to participate in support vector machine combination, the results show that the proposed method The prediction accuracy is not only higher than the single model, but also higher than the linear combination forecasting model and neural network forecasting model.