针对风电场短期风速的预测提出一种基于小波变换的组合预测方法。首先利用Mallat 算法对短期风速时间序列进行db3 小波三层分解与重构，得到短期风速时间序列的近似分量和细节分量。针对近似分量和细节分量的不同特性，对近似分量利用粒子群算法优化的最小二乘支持向量机进行预测，对细节分量利用自回归求和滑动平均模型进行预测。最后各预测模型预测值组合叠加得到最终的短期风速预测值。仿真结果表明该方法具有较高的预测准确度。-In order to improve short-term wind speed prediction accuracy of wind farms，a combined prediction method based on the wavelet transform is proposed. Firstly，the db3 wavelet is used for three-layer decomposition and reconstruction for short-term wind speed time series through Mallat algorithm. The approximation components and the detail components of the short-term wind speed are then obtained. Next，according to the different characteristics of these components，the least square support vector machine optimized

by particle swarm algorithm and the autoregressive integrated moving average model are adopted as the predictivemodels for the approximate components and the detail components respectively. Then，the final predictive value of the short-term wind speed is obtained by the combination of the two components. The simulation results indicate that higher accuracy can be obtained in this prediction method.