摘要
为提高“煤改电”地区短期负荷预测水平,本文基于北京市大兴区“煤改电”工程,探索利用神经网络算法对“煤改电”地区短期负荷进行预测。本文首先研究了“煤改电”地区负荷的年周期、周周期以及日周期负荷特性,并对负荷预测进行分类,分析得出了负荷预测的主要影响因素,明确了负荷预测的步骤及误差分析方法。其次,本文研究了BP神经网络的构成和运算过程,分析了历史数据处理方法,建立了基于BP神经网络的“煤改电”地区短期负荷预测模型,并对短期负荷预测模型进行检验。最后,为进一步提高预测效果,本文研究利用粒子群算法和列文伯格-马夸尔特算法对神经网络进行优化改进,建立了基于粒子群算法优化的BP神经网络负荷预测模型,满足了预测目标精度要求。
In order to improve the level of short-term load forecasting in the"coal-to-electricity"area,this article explores the use of neural network algorithms to forecast the short-term load of the"coal-to-electricity"area based on the"coal-to-electricity"project in Daxing District,Beijing.This article first studies the load characteristics of the annual cycle,weekly cycle and daily cycle load in the"coal-to-electricity"area,classifies the load forecast,analyzes the main influencing factors of the load forecast,and clarifies the steps and error analysis of the load forecast method.Secondly,this article studies the composition and calculation process of BP neural network,analyzes the historical data processing method,establishes a short-term load forecasting model of"coal-to-electricity"area based on BP neural network,and tests the short-term load forecasting model.Finally,in order to further improve the forecasting effect,this article studies the use of particle swarm algorithm and Levenberg-Marquardt algorithm to optimize and improve the neural network,and establishes a BP neural network load forecasting model based on particle swarm algorithm optimization,which meets the requirements of forecast target accuracy.
作者
赵迪
孟静
李志
张恩领
卢瑾
陆子昂
ZHAO Di;MENG Jing;LI Zhi;ZHANG Enling;LU Jin;LU Ziang(Daxing Power Supply Branch of State Grid Beijing Electric Power Company,Beijing 102600,Beijing,China)
出处
《电力大数据》
2021年第1期40-47,共8页
Power Systems and Big Data
关键词
煤改电
负荷预测
神经网络
列文伯格-马夸尔特算法
粒子群算法
"coal-to-electricity"area
load forecast
neural network
levenberg-marquardt algorithm
particle swarm optimization
