基于人工神经网络的风电功率预测04267.pdfVIP

第28 卷 第34 期 中 国 电 机 工 程 学 报 Vol.28 No.34 Dec. 5, 2008 118 2008 年12 月5 日 Proceedings of the CSEE ©2008 Chin.Soc.for Elec.Eng. 0258-8013 (2008) 34-0118-06 TM 743 A 470 40 文章编号： 中图分类号： 文献标志码： 学科分类号： ⋅ 基于人工神经网络的风电功率预测 范高锋，王伟胜，刘纯，戴慧珠 ( 中国电力科学研究院，北京市 海淀区 100192) Wind Power Prediction Based on Artificial Neural Network FAN Gao-feng, WANG Wei-sheng, LIU Chun, DAI Hui-zhu (China Electric Power Research Institute, Haidian District, Beijing 100192, China) ABSTRACT: Wind power prediction is important to the 些新问题，其中很重要的一方面是对电力系统运行 operation of power system with comparatively large mount of 调度的影响[1-4] 。风电功率预测对电力系统的功率平 wind power. The wind power prediction methods were 衡和经济调度具有非常重要的意义。国外风电装机 classified into several kinds. An artificial neural network (ANN) 容量较大的国家都进行了风电功率预测系统的研 model for wind power prediction was constructed according to 究与开发。风电功率预测方法根据预测的物理量来 the wind power influence factors. Then the impacts of real time 分类，可以分为2 类：第 1 类为对风速的预测，然 measured power and the atmospheric data at different heights 后根据风电机组或风电场的功率曲线得到风电场 on prediction results were analyzed. Besides, another ANN model for error band prediction was also built. The results 功率输出；第2 类为直接预测风电场的输出功率。 indicate that the ANN structure and the training sample have 根据所采用的数学模型不同可分为持续预测法、自 some impact on the prediction precision. The real time 回归滑动平均(auto regressive moving average, measured power as input will improve the precision of 30 min ARMA)模型法、卡尔曼滤波法和智能方法等。持续 ahead prediction, howe