基于小波神经网络的电力负荷预测方法  

Study of power system load forecast based on wavelet neural networks

作　　者：张大海[1] 江世芳[1] 毕研秋[1] 邹贵彬[1] 

机构地区：[1]山东大学电气学院,山东济南250061

出　　处：《电力自动化设备》2003年第8期29-32,共4页Electric Power Automation Equipment

摘　　要：分析了小波神经网络的特点,研究了在电力负荷预测中小波神经网络存在的优缺点及适用范围。通过对小波神经网络和BP神经网络的结构和算法进行理论分析,并对实际电力负荷预测算例进行对比研究,指出小波神经网络本身适合对波动性的信号进行预测,而且在神经网络节点数目相同的情况下,小波神经网络比BP神经网络具有更高的预测精度,因此采用小波神经网络有利于减少隐节点数目。还指出由于当前的连续小波神经网络主要使用传统BP神经网络的随机初始化方法和基于梯度的训练算法,因此存在收敛性差的缺点。The characters of WNN(Wavelet Neural Network)are analyzed and the advantages and disadvantages of its applications in power system load forecast are studied.By theoretic analyzing the network structures and algorithms of WNN and BP Net and comparing the forecast results of power load,it is pointed out that WNN is suitable for forecasting the variable signals.When they have same number of network node,WNN is better than BP NET in forecast accuracy.Therefore,WNN may be applied to reduce the number of hidden node.It is also indicated that current WNN has a poor convergence performance because of adopting the random initialization method and gradient training algorithm of traditional BP NET.

关 键 词：负荷预测 小波理论 小波神经网络 BP神经网络 

分 类 号：TM715[电气工程—电力系统及自动化]