摘要
针对核极限学习机(kernel extreme learning machine,KELM)单一预测模型不稳定以及预测结果不准确,提出了一种变分模态分解(variational mode decomposition,VMD)与麻雀搜索算法(sparrow search algorithm,SSA)优化的混合核极限学习机(hybrid extreme learning machine,HKELM)模型。首先把预处理后的负荷序列依据变分模态技术分解为若干相对平稳的模态分量,然后同时对每个模态分量建立VMD-SSA-HKELM预测模型;再将负荷数据划分训练集和测试集;依据训练集分别用SSA算法优化HKELM的参数,将测试集代入每个模型,所测的结果叠加得出最终预测值。该模型采用麻雀算法优化的混合核极限学习机,使其在不同的参数下有良好的局部搜索能力,且能增强全局搜索能力。仿真结果表明,VMD-SSA-HKELM模型预测精度接近98.5%,为超短期负荷预测及电力系统稳定运行提供了决策的支持。
For kernel extreme learning machine(KELM)unstable single prediction model and inaccurate prediction results,this paper presents a variational mode decomposition(VMD)and the sparrow search algorithm(SSA)optimized hybrid extreme learning machine(HKELM)model.The preprocessed load sequence is decomposed into several relatively stable modal components according to variational modal technology,and then establish VMD-SSA-HKELM prediction model for each modal component;then the load data is divided into the training set and test set;optimize the parameters of HKELM by SSA algorithm,then substitute the test set into each model,and the measured results are accumulated to obtain the final prediction value.The proposed model uses a mixed kernel extreme learning machine optimized by the sparrow algorithm to have good local search ability under different parameters,and can enhance the global search capability.The simulation results show that the prediction accuracy of VMD-SSA-HKELM model proposed is close to 98.5%,which provides decision support for ultra-short-term load prediction and stable operation of power system.
作者
郭建帅
崔双喜
郭建斌
姚岱伟
孙冠岳
Guo Jianshuai;Cui Shuangxi;Guo Jianbin;Yao Daiwei;Sun Guanyue(School of Electrical Engineering,Xinjiang University,Urumqi 830047,China;Skills Training Center of State Grid Shanxi Electric Power Company,Taiyuan 030021,China)
出处
《国外电子测量技术》
北大核心
2022年第6期105-111,共7页
Foreign Electronic Measurement Technology
关键词
变分模态分解
麻雀搜索算法
混合核极限学习机
组合预测
超短期负荷预测
variational mode decomposition
sparrow search algorithm
hybrid kernel extreme learning machine
combination prediction
ultra-short-term load prediction
