基于VMD-PSO-多核极限学习机的短期负荷预测  

Short-term Load Forecasting Based on VMD-PSO-MKELM Method

作　　者：吴松梅 蒋建东[1] 燕跃豪 鲍薇 WU Songmei;JIANG Jiandong;YAN Yuehao;BAO Wei(School of Electrical Engineering,Zhengzhou University,Zhengzhou 450000,China;Zhengzhou Power Supply Company,State Grid Henan Electric Power Company,Zhengzhou 450000,China)

机构地区：[1]郑州大学电气工程学院,郑州450000 [2]国网河南省电力公司郑州供电公司,郑州450000

出　　处：《电力系统及其自动化学报》2022年第5期18-25,共8页Proceedings of the CSU-EPSA

基　　金：国家自然科学基金资助项目(51507155);河南省研究生教育改革与质量提升工程项目(YJS2021JD02)。

摘　　要：为提高短期负荷预测精度,解决核极限学习机单一核函数难以适应负荷多数据特征的问题,提出了一种基于变分模态分解与粒子群优化的多核极限学习机模型。该模型采用变分模态分解技术将原始负荷序列分解为具有不同特征频率的子序列,并对每个子序列建立预测模型。负荷预测模型采用粒子群优化的多核极限学习机,其使用混合核函数代替单一的核函数,使其在不同的参数下不仅有良好的局部搜索能力,同时也加强了全局搜索能力。实验表明,该模型拥有更好的回归精度和泛化能力,能够得到更精确的预测结果。To improve the accuracy of short-term load forecasting(STLF)and solve the problem that the single-kernel function of kernel extreme learning machine(KELM)is difficult to adapt to the characteristics of multiple data sam⁃ples,a multi-kernel extreme learning machine(MKELM)model is proposed in this paper,which is based on variation⁃al modal decomposition(VMD)and particle swarm optimization(PSO).This model uses the VMD technology to decom⁃pose the original load sequence into a series of sub-sequences with different characteristic frequencies,and a forecast⁃ing model is established for each sub-sequence.The load forecasting model adopts the MKELM optimized by PSO,and uses a hybrid-kernel function to replace the single-kernel function,so that it not only has a good local search capability un⁃der different parameters,but also strengthens its global search capability.Experimental results show that this model has a higher regression accuracy and a better generalization capability,which can obtain more accurate prediction results.

关 键 词：变分模态分解 粒子群优化算法 核函数 多核极限学习机 短期负荷预测 

分 类 号：TM715[电气工程—电力系统及自动化]