基于神经网络方法的大型电网短期负荷预报  

China Northeast Power Network Short-term Load Forecasting Based on Neural Network

作　　者：赵希人[1] 彭秀艳[1] 姜广宇[1] 

机构地区：[1]哈尔滨工程大学自动化学院,哈尔滨150001

出　　处：《系统仿真学报》2006年第6期1677-1680,共4页Journal of System Simulation

摘　　要：电力系统负荷预报研究现状,介绍了神经网络方法应用于电力系统短期负荷预报的可行性及存在的问题。详细讨论了应用BP神经网络、共轭梯度算法改进BP神经网络方法进行电力系统短期负荷预报的算法,及在预报过程中对电网负荷数据进行预处理方法。分别应用二种方法对东北电力系统进行了72小时短期负荷预报仿真。仿真结果表明,BP神经网络训练时间长,预报精度低;而共轭梯度算法改进BP神经网络算法训练步数大大减小,缩短了网络训练时间,而且提高了预报精度。该方法可行,可用于电力系统短期负荷在线预报。Harbin Engineedn8 University, Harbin 150001, China） Abstract： The search status of power network short-term load forecasting was introduced first, and the feasibility and problem that neural network was used in power network short-term load forecasting were introduced. The algorithms were discussed detailedly, which used BP neural network and improved BP neural network with conjugate gradient algorithm forecast power network short-term load. And the preprocessing method of power network history data was given. Then the simulation results of two kinds of forecasting method were given for 72 hours power network load of northeast of China. According to the forecasting simulation results, the conclusions are given that BP neural network training time is longer and forecasting precision is lower, but the improved BP neural network algorithm with Conjugate gradient algorithm makes the training times shorter, and it enhances forecasting precision. The calculation results show that the approaches of the improved BP neural network with conjugate gradient algorithm is feasible and can be used to forecast on line.

关 键 词：短期负荷预报 BP神经网络 共轭梯度法 在线预报 仿真 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]