基于GA-BP-LSTM模型的配电台区低电压预测  

Low Voltage Prediction in Distribution Station Area Based on GA-BP-LSTM Model

作　　者：魏金柱 田施兰 陈雯晨 况露 马志鹏[2] WEI Jinzhu;TIAN Shilan;CHEN Wenchen;KUANG Lu;MA Zhipeng(Qianjiang Power Supply Branch of State Grid Chongqing Electric Power Company,Chongqing 409000,China;Chongqing University of Technology,Chongqing 400054,China)

机构地区：[1]国网重庆市电力公司黔江供电分公司,重庆409000 [2]重庆理工大学,重庆400054

出　　处：《电工技术》2025年第1期15-20,共6页Electric Engineering

基　　金：国家电网公司科技项目(编号SGCQQJ00FJJS2400-218)。

摘　　要：随着电力需求的不断增长和电力负荷的复杂变化,台区低电压问题已成为影响供电质量和用户体验的关键因素之一。在这种背景下,对台区低电压进行精准预测,不仅是电力系统优化的迫切需要,更是确保供电质量、提升用户满意度的重要手段,对于解决低电压问题具有重要意义。构建了基于GA-BP神经网络的台区负荷预测模型,该模型考虑了温度、湿度、日期类型、一天不同时刻等多种影响因素,通过神经网络的自学习和自适应能力,借助遗传算法的全局最优搜索方面的卓越性能优化BP神经网络的权值与阈值,改善预测性能。由于影响电压的主要因素为负荷,LSTM算法通过引入门控循环单元(包括遗忘门、输入门、输出门等),能够有效地记住并处理负荷数据中的依赖关系,并建立LSTM模型,将已预测到的负荷数据作为输入来实现对电压的预测。利用MATLAB软件对上述组合模型进行编程并结合实例进行分析,结果表明该组合模型能够实现对台区低电压的预测。With the continuous growth of power demand and complex changes in power load,the problem of low voltage in the station area has become one of the key factors affecting the quality of power supply and user experience.In this context,accurate prediction of low voltage in the station area is not only an urgent need for power system optimization,but also an important means to ensure the quality of power supply and improve user satisfaction,which is of great significance for solving the problem of low voltage.In this paper,a load prediction model based on GA-BP neural network is constructed,which considers various influencing factors such as temperature,humidity,date type,and different times of the day,and optimizes the weights and thresholds of BP neural network through the self-learning and self-adaptation ability of the neural network,and optimizes the weights and thresholds of the BP neural network with the help of the excellent performance of the genetic algorithm in terms of global optimal search,so as to improve the prediction performance.Since the main factor affecting the voltage is the load,the LSTM algorithm can effectively remember and process the dependencies in the load data by introducing the gated recurrent unit(including the forgetting gate,input gate,output gate,etc.),and the predicted load data is used as input to realize the voltage prediction by establishing the LSTM model.The above combined model is programmed by MATLAB software and combined with examples,and the simulation results show that the combined model can predict the low voltage in the station area.

关 键 词：配电台区 负荷预测 电压预测 BP神经网络 长短期神经网络 

分 类 号：TM713[电气工程—电力系统及自动化]