为提高光伏出力预测精度，针对光伏出力序列非线性、非平稳特点，利用具有平稳化处理技术的ESMD法，将光伏出力序列分解为各模态分量和趋势余项，并将高频模态1,2进一步分解，结合SVM模型分别预测各平稳序列，建立具有“分解-预测-重构”特征的ESMD-SVM组合模型，以某光伏发电厂为例验证模型有效性。结果表明：ESMD-SVM组合模型预报精度大于单一模型，与单一SVM预测模型相比，均方根误差降低了84.81%、平均绝对误差降低了81.27%，决定系数提高了7.42%。

In order to improve the precision of photovoltaic output prediction,according to the nonlinear and nonstationary characteristics of photovoltaic output sequence,the photovoltaic output sequence was decomposed into each modal component and trend remainder by using extremepoint symmetric mode decomposition (ESMD) method with stabilization processing technology. The high frequency modes Ⅰ and Ⅱ were further decomposed,and the stationary sequences were predicted respectively combined with support vector machine (SVM) model. The ESMDSVM combined model with the characteristics of "decompositionpredictionreconstruction" was established,and the the model validity was verified by taking a photovoltaic power plant as example. The results show that the prediction accuracy of the ESMDSVM combined model is higher than that of the single model,the root mean square error (RMSE) is reduced by 84.81%,the mean absolute error (MAE) is reduced by 81.27%,and the determination coefficient is increased by 7.42% compared with the single SVM prediction model.

TM615

电网安全与节能国家重点实验室开放基金(YDB51202101236)