文献类型：期刊文献

中文题名：时序神经网络算法的短期风速预测研究

英文题名：On Short-Time Wind Speed Prediction Based on Time-Series Neural Network Algorithm

作者：梁岚珍[1];邵璠[2]

第一作者：梁岚珍

机构：[1]北京联合大学自动化学院;[2]新疆大学电气工程学院

第一机构：北京联合大学城市轨道交通与物流学院

年份：2011

卷号：18

期号：1

起止页码：43-45

中文期刊名：控制工程

外文期刊名：Control Engineering of China

收录：CSTPCD;;北大核心:【北大核心2008】；CSCD:【CSCD2011_2012】；

基金：北京市教委科研计划项目(KM200911417012)

语种：中文

中文关键词：风速;预测;时间序列法;BP神经网络

外文关键词：wind speed; prediction; time-series method; back-propagation network

摘要：采用神经网络对风速进行短期预测,研究BP型短期风速预测网络中BP算法、BP网络构建以及网络训练方法。结合时间序列法和神经网络法提出了时序神经网络预测方法,对短期风速预测网络中输入变量数量和隐含层节点数量的选择方法进行了探讨。仿真实验结果表明,时序神经网络法建立的网络,训练时间明显缩短,网络输出的预测值与真实的观察值之间没有出现过大偏差,对新训练样本集拟合较好,预测精度有所提高,并在一定程度上克服了基本BP网络收敛速度慢以及易陷入局部最小等问题。
The short-time wind speed prediction problem is discussed by using the neural network.The algorithm,the construction,and the training and prediction way of BP network are studied.The wind speed short-time prediction neural network algorithm is presented.The time-series neural network prediction is proposed based the method of time-series and network algorithm.The way to choose the numbers of input variables and implicit layer node is given.The simulate result shows that the network based on time-series neural network prediction has the disadvantages of much shorter training time,small error between predicted data and real data and goodness of fitting and predicting accuracy.The predicting method overcomes the disadvantages of slow convergence velocity and local least of BP network.