改进LSTM架构下基于短期负荷预测的智能电网分布式调度策略  

Intelligent Grid Distributed Dispatching Strategy Based on Short-term Load Forecasting under Improved LSTM Architecture

作　　者：李彪 魏飞 蒋德玉 张恒 曲小康 LI Biao;WEI Fei;JIANG Deyu;ZHANG Heng;QU Xiaokang(Linyi Power Supply Company of State Grid Shandong Power Supply Company,Linyi 276003,China)

机构地区：[1]国网山东省电力公司临沂供电公司,山东临沂276003

出　　处：《微型电脑应用》2025年第2期38-41,共4页Microcomputer Applications

基　　金：国网山东省电力公司科技项目(2021A-028)。

摘　　要：针对传统短期电网负荷预测算法无法对海量数据进行处理的问题,基于改进LSTM模型提出一种预测模型。所提模型采用Bi-LSTM来提升单向LSTM的数据处理耦合性,使模型具备对依据历史与未来数据的耦合特征进行特征提取的能力。通过注意力机制扩增模型的通道数,增强数据特征的提取能力,使用麻雀搜索算法对模型参数加以优化,再根据短期负荷预测结果制定用电策略。在实验测试中,所提算法的预测结果更贴合实际曲线,分布式电能调度策略也可有效降低用电负荷,且RMSE、MAE、MAPE误差指标与原算法相比降低了0.144、0.070和0.52个百分点,证明了所提算法性能良好,能够投入实际的工程应用。Aimed at the disadvantage that the traditional short-term load forecasting algorithm cannot process massive data,this paper proposes a short-term load forecasting model based on LSTM model.The proposed model uses bidirectional LSTM to improve the one-way coupling,so that the model has the ability to extract the coupling features of historical data and future data.The number of channels of the model is expanded through the attention mechanism,thus the ability of data feature extraction is enhanced.The sparrow search algorithm is used to optimize the model parameters,and the power consumption strategy is formulated according to the short-term load forecasting results.In the experimental test,the prediction results of the proposed algorithm are more consistent with the actual curves,the distributed power scheduling strategy can effectively reduce the power load,and the RMSE and MAPE error indicators are reduced by 0.144,0.070 and 0.52 percentage points compared with the original algorithm,which proves that the proposed algorithm has good performance and can be applied to actual engineering projects.

关 键 词：长短时记忆网络 麻雀搜索算法 注意力机制 电网负荷预测 调度控制 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN929.5[自动化与计算机技术—计算机科学与技术]