基于主成分—遗传神经网络的短期风电功率预测  

Short-term wind power prediction based onprincipal component analysis and genetic neural network

作　　者：罗毅[1] 刘峰[1] 刘向杰[1] 

机构地区：[1]华北电力大学控制与计算机工程学院,北京102206

出　　处：《电力系统保护与控制》2012年第23期47-53,共7页Power System Protection and Control

基　　金：国家自然科学基金项目(60974051);北京市自然科学基金项目(4122071)~~

摘　　要：短期风电功率预测对接入大量风电的电力系统运行具有重要的意义,建立了基于主成分分析与遗传神经网络相结合的短期风电功率预测模型。该模型先对原始输入数据进行主成分分析,分析结果作为神经网络预测模型的输入;为克服BP神经网络训练时间长、易陷入局部极小值的的缺陷,采用遗传算法优化神经网络的初始权值和阈值,并使用Levenberg-Marquardt算法对网络权值和阈值进行细化训练。经某风电场实际数据验证,与GA神经网络模型、PCA-LM神经网络模型相比,预测精度明显提高,为短期风电功率预测提供了一种有效的方法。Short-term wind power prediction is important to the operation of power system with comparatively large amount of wind power, a short-circiut wind power predicting model based on principal component analysis （PCA） method and genetic neural network is proposed. PCA is applied to process original input data, the principal components are used as input data for neural network. In order to solve the problems of slow convergence speed and being easy to fall into local minimum of BP neural network, genetic algorithm（GA） is used to make a thorough searching for the initial weights and thresholds, and the Levenberg-Marquardt （L-M） method is used to finely train the network. Based on the actual data of a wind farm, the forecasting results by the proposed method is more precise than those by GA neural network model and PCA-LM neural network model, providing an effective way to forecast short-term wind power.

关 键 词：风电功率 神经网络 遗传算法 主成分分析 短期预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TM614[自动化与计算机技术—控制科学与工程]