摘要
神经网络用于火电厂时间序列预报无需作平稳性假设。它从序列样本中学习序列本身的内在规律，从而建立正确的火电厂时间序列模型。神经网络将寻求序列规律的过程转化为Ｒｎ →Ｒｍ 逼近的非线性映射的非线性优化问题，用经过改进的静态ＢＰ算法可以取得较为满意的结果。适当增加输入单元的历史序列样本，可以得到对序列更为精确的预报。
Neural network applied in power plant forecasing time sequence does not need stability hypothesizing.It studies interior laws of sequence from stylebooks of sequence,then builds a proper plant time sequence model.With neural network,the process to seek sequence laws is changed into a non linear optimizing problem of non linear mapping by R n→R m approach.The stable BP algorithmic means improved can obtain satisfying results.Forecasting sequence is more precise if historical input is increased properly.Examples have indicated that neural network can forecast plant time sequence better.Figs 5,tables 3 and refs 6.
出处
《动力工程》
EI
CAS
CSCD
2000年第1期554-557,共4页
Power Engineering
关键词
火电厂
神经网络
时间序列
模型
预报
fired power plant
neural network
time sequence
model
forecasting
