基于MVMCGS组合模型的短期电力负荷预测研究  

Study on short-term electric load forecasting based on MVMCGS combination modeling

作　　者：陈金龙 吴斌(指导)[1] 孙丽 CHEN Jinlong;WU Bin;SUNLi(School of Business,Shanghai DianJi University,Shanghai 201306,China)

机构地区：[1]上海电机学院商学院,上海201306

出　　处：《上海电机学院学报》2025年第1期14-20,共7页Journal of Shanghai Dianji University

基　　金：教育部人文社会科学青年基金项目(20YJCZH027)。

摘　　要：随着电力系统的不断更新,其负荷波动性及非线性特征变得难以预测,进而导致负荷预测精度不高。针对这一问题,提出了一种基于多策略鲸鱼优化算法(MSWOA)、多尺度样本熵(MSE)优化的变分模态分解(VMD)与冠豪猪优化(CPO)算法、自注意力机制优化的门控循环单元(GRU)相结合的组合预测模型,即MVMCGS组合预测模型。首先,利用复合混沌映射、自适应权重策略、高斯变异结合淘金优化算法增强鲸鱼优化算法的局部和全局搜索能力,得到MSWOA;然后,结合MSE优化VMD,提高分解效果;最后,采用CPO优化GRU模型参数和自注意力机制优化GRU模型的权重分配,将分解后的子序列导入模型中,提高模型预测精度。实验表明:该模型的预测能力表现出色,其预测精度优于单一模型和其他相关混合模型,验证了该模型的有效性。As power systems continue to evolve,the load volatility and nonlinearity become increasingly difficult to predict,resulting in poor forecasting accuracy.To address this issue,a combined prediction model,named the MVMCGS model,is proposed.This model integrates the multi-strategy whale optimization algorithm(MSWOA),multiscale sample entropy(MSE)-optimized variational mode decomposition(VMD),crested porcupine optimization(CPO),and selfattention optimized gate recurrent unit(GRU).Firstly,the MSWOA is derived by employing composite chaotic mapping,adaptive weighting strategies,gaussian mutation,and integrating the gold rush algorithm to enhance the local and global search capabilities of the whale optimization algorithm.This is followed by using MSE to optimize the VMD to improve its decomposition performance.Next,the decomposed subsequences are input into the GRU model,where the CPO algorithm is used to optimize the GRU parameters, and the self-attention mechanism is employed tooptimize the weight distribution in the GRU model. This approach improves the overall prediction accuracyof the model. Experimental results show that the MVMCGS model demonstrates excellent predictionperformance, outperforming both single models and other hybrid models in terms of predictionaccuracy, thus validating the model's effectiveness.

关 键 词：多策略鲸鱼优化算法 多尺度样本熵 自注意力机制 组合预测模型 

分 类 号：TM715[电气工程—电力系统及自动化]