基于MA-LSTM的短期负荷预测方法  

MA-LSTM-based Method of Predicting Short-term Load

作　　者：尹辉彦 常勇 杨海兰[1] 武文成 巩锴 YIN Huiyan;CHANG Yong;YANG Hailan;WU Wencheng;GONG Kai(Gansu Polytechnic College of Animal Husbandry and Engineering,Wuwei 733006,China;Lanzhou Petrochemical University of Vocational Technology,Lanzhou 730060,China;Gansu Transmission and Transformation Engineering Co.,Ltd.,Lanzhou 730050,China)

机构地区：[1]甘肃畜牧工程职业技术学院,甘肃武威733006 [2]兰州石化职业技术大学,甘肃兰州730060 [3]甘肃送变电工程有限公司,甘肃兰州730050

出　　处：《电工技术》2023年第21期15-19,共5页Electric Engineering

基　　金：武威市2021年度市列科技计划项目(编号WW2101005);2021年国家级大学生创新创业训练计划项目(编号202113955001)。

摘　　要：精准的预测电力系统短期负荷对电力系统智能化和可靠运行有重要意义。为了提高负荷预测的精度,采用了一种基于蜉蝣优化算法和长短期记忆神经网络的短期负荷预测方法。将影响负荷的温度、日期类型、湿度作为输入特征;对长短期记忆神经网络中的参数使用蜉蝣算法不断地优化以确定最优参数;最后建立MA-LSTM模型对短期负荷进行预测。算例结果表明,和BP、LSTM、PSO-LSTM及SSA-LSTM方法相比,所提方法具有更高的预测精度,为电网安全稳定运行提供了有力保障。Accurate short-term load prediction of power systems is of great significance for intelligent and reliable operation of power systems.A short-term load prediction method based on mayfly optimization algorithm(MA)and long short term memory neural network(LSTM)is proposed,fully considering the characteristics of power load data such as timeliness,nonlinearity,and periodicity.Taking temperature,date type,and humidity that affect the load as input characteristics,the parameters in long-term and short-term memory neural network are continuously optimized using the mayfly algorithm to determine the optimal parameters.Finally,a MA-LSTM model is established to predict short-term load.The results of an example show that the proposed method has higher prediction accuracy compared to other independent and integrated methods,benefiting safe and stable operation of power grid.

关 键 词：短期负荷 蜉蝣优化算法 长短期记忆神经网络 预测 

分 类 号：TM714[电气工程—电力系统及自动化]