基于混沌理论与神经网络的短期风速滚动预测  

Short-Term Wind Speed Rolling Forecast Based on Chaos Theory and Neural Network

作　　者：华琦[1] 王雷[2] 陆金桂[2] 杜杰[3] 

机构地区：[1]南京工业大学自动化与电气工程学院,南京211008 [2]南京工业大学机械与动力工程学院,南京211008 [3]南京信息工程大学计算机与软件学院,南京210044

出　　处：《电测与仪表》2013年第7期16-20,共5页Electrical Measurement & Instrumentation

基　　金：国家自然科学基金资助项目(61103142)

摘　　要：风的演变具有混沌特性,论文以混沌相空间重构理论为基础,应用神经网络技术进行风速短期预测。采用互信息法和饱和关联维数法分别计算最佳的延迟时间和嵌入维数,将原风速时间序列数据进行相空间重构,并用wolf法计算最大Lyapunov指数对时间序列的混沌特性进行判别。论文利用神经网络建立风速预测模型,并进行多步滚动式短期风速预测。论文还进行了仿真实验,实验结果初步验证了基于混沌理论的神经网络预测模型在风速短步长预测方面的效果。This paper is involved in the wind speed forecast model based on chaos theory and artificial neural network.The delay time and the embedding dimension are obtained with adopting the mutual information method and Grassberger-Procaccia algorithm.When the phase space of the original wind speed time series data is reconstructed,the new space can reflect the characteristics of wind speed with which the largest Lyapunov exponent is calculated with the wolf method.So the multi-step rolling forecast model using neural network is gotten in the paper.In the end,the experiment result of the simulation demonstrates that this model is helpful for improving accurate results.

关 键 词：短期风速预测 滚动式多步预测 混沌相空间重构 神经网络 

分 类 号：TM614[电气工程—电力系统及自动化]