基于NARX神经网络的光伏发电功率预测研究  

Generated Power Forecasting for Photovoltaic System Based on NARX Neural Network

作　　者：付青[1] 单英浩 朱昌亚 

机构地区：[1]中山大学物理科学与工程技术学院,广东广州510006 [2]天宝电子(惠州)有限公司,广东惠州516005

出　　处：《电气传动》2016年第4期42-45,共4页Electric Drive

基　　金：广东省科技计划(2013B010405009);珠海市战略性新兴产业重大专项(2014D0601990002);广东省产学研合作项目(2014B090903009)

摘　　要：为了较准确地预测光伏发电系统的发电功率,建立了动态神经网络预测模型。该模型采用有外部输入的非线性自回归神经网络(NARX)结构,考虑太阳能辐射量和电池板温度对光伏发电功率的影响,利用NARX神经网络强大的非线性映射和泛化能力,进行了发电功率的预测。结果表明,利用NARX神经网络预测光伏发电功率是可行的,并且与传统BP神经网络相比,具有良好的适应性和预测精度。To forecast the generated power of photovoltaic （PV） system accurately, the model of generated power forecasting of PV system based on dynamic neural network was built. Nonlinear auto-regressive with external/ exogenous inputs （NARX） neural network topology was adopted. Solar irradiation quantity and temperature of PV modules were considered to forecast generated power by using the powerful ability of nonlinear mapping and generalization of NARX. The results show that NARX neural network is appropriate to the prediction of PV system generated power and this method has better adaptability and prediction accuracy than traditional BP network.

关 键 词：光伏系统 有外部输入的非线性自回归 神经网络 发电功率预测 

分 类 号：TM615[电气工程—电力系统及自动化]