针对短期风电功率预测，提出一种基于经验小波变换预处理的核极限学习机组合预测方法。首先采用 EWT 对风电场实测风速数据进行自适应分解并提取具有傅立叶紧支撑的模态信号分量，针对每个分量分别构建 KELM 预测模型，最后对各个预测模型的输出进行叠加得到风速预测值并根据风电场风功特性曲线可得对应风电功率预测值。(Aiming at short-term wind power prediction, a kernel-based learning machine combination prediction method based on empirical wavelet transform preprocessing is proposed. Firstly, the EWT is used to adaptively decompose the measured wind speed data of the wind farm and extract the modal signal components with Fourier tight support. The KELM prediction model is constructed for each component. Finally, the output of each prediction model is superimposed to obtain the wind speed prediction value. The wind power characteristic curve of the wind farm can be obtained corresponding to the predicted wind power.)