基于EEMD-EN-SVR的短期负荷预测方法

Short-term Load Forecasting Method Based on EEMD-EN-SVR

支持向量回归(SVR)应用于短期负荷预测时,出现训练时间变长、结果精度下降现象的主要原因是由于输入中存在冗余不相关特征.考虑到弹性网回归(EN)能够有效剔除冗余不相关特征的特点,构建基于EEMD-EN-SVR的短期负荷预测模型.该模型采用集合经验模态分解(EEMD)对负荷序列进行分解并提取用电特征,使用EN方法进行特征选择,筛选出冗余和不相关特征,获得最佳特征集,最终利用粒子群算法优化的支持向量回归(PSO-SVR)对短期负荷进行预测.通过某地区的真实用电数据进行实验分析,并与文中所提到的其他模型进行比较,结果表明,所提方法的预测精度和鲁棒性较好.

When support vector regression(SVR)is applied to short-term load forecasting,the main reason for the long training time and the decrease of the accuracy of the results is due to the existence of redundant or irrelevant features in the input.Considering that elastic network regression(EN)can effectively eliminate the characteristics of redundant or unrelated features,a short-term load forecasting model based on EEMDEN-SVR is constructed.In this model,the empirical model decomposition(EEMD)is used to decompose the load sequence and extract the electrical characteristics.Then the EN method is used to select features,and the redundant and irrelevant features are selected to obtain the best feature set.Finally,particle swarm-opti⁃mization supported vector regression(PSO-SVR)is used to predict short-term load.The experimental analy⁃sis is carried out through the real electricity consumption data of a certain area,and compared with other models mentioned in this paper,the results show that the prediction accuracy and robustness of the proposed method are better.