基于主成分分析与前向反馈传播神经网络的风电场输出功率预测.pdfVIP

第 35 卷 第 3 期 电 网 技 术 Vol. 35 No. 3 2011 年 3 月 Power System Technology Mar. 2011 文章编号：1000-3673 （2011 ）03-0183-05 中图分类号：TM 614 文献标志码：A 学科代码：470·4047 基于主成分分析与前向反馈传播神经网络的 风电场输出功率预测 张明理，杨晓亮，滕云，徐建源，林莘 （沈阳工业大学电气工程学院，辽宁省 沈阳市 110870 ） Forecasting of Wind Farm Generation Output Based on Principal Component Analysis and Back Propagation Neural Network ZHANG Mingli, YANG Xiaoliang, TENG Yun, XU Jianyuan, LIN Xin (School of Electrical Engineering, Shenyang University of Technology, Shenyang 110870, Liaoning Province, China) ABSTRACT: To remedy the defects existing in the 举措。但是风电具有间歇性和随机性的固有缺点， forecasting of wind farm generation output by tranditional 随着大量风力发电接入电网，势必会对电力系统的 singular forecasting method, a forecasting method based on 安全、稳定运行以及保证电能质量带来严峻的挑 principal component analylsis (PCA) and back propagation 战。因此风电场输出功率预测成为解决这一问题的 neural netwrok (BPNN) is proposed. Firstly, the samples are chosen by K-S method; then the effective information in the 有效途径之一[1-4] 。 chosen samples is extracted by PCA and the principal 目前风电场输出功率主要的预测方法根据所 components are solved; and then the BPNN is constructed to 采用数学模型的不同可分为持续预测法、自回归滑 forecast the wind farm generation output. Calculation results 动平均模型法、卡尔曼滤波法和智能方法等。其中 show that after the processing by PCA the correlativity among 智能方法则主要采用基于数值天气预报的预测模 input factors in BPNN model is eliminated as well as the structure of the BPNN is simplified, thus the BPNN converges 型，分为物理模型和统计模型(如神经网络方法、模 faster. The forecasting results by