基于天气类型聚类和LS-SVM的光伏出力预测  

Photovoltaic System Output Power Forecasting Based on Weather Type Clustering and LS-SVM

作　　者：张华彬[1] 杨明玉[1] 

机构地区：[1]华北电力大学电气与电子工程学院,河北保定071003

出　　处：《电力科学与工程》2014年第10期42-47,共6页Electric Power Science and Engineering

摘　　要：分析了影响光伏出力的气象因素,结合光伏系统实际运行数据和气象信息,提出一种基于天气类型聚类和LS-SVM的光伏出力预测模型。选取太阳辐照时间、温度、相对湿度等作为气象特征向量,通过计算各向量的加权欧氏距离,筛选出最佳聚类集合,确定训练样本,使样本数据能更好地反映待预测日的实际气象信息。取最佳聚类日气象特征、相应光伏出力及待预测日气象特征输入训练好的LS-SVM模型,输出为待预测日对应时刻的光伏出力。最后通过实际算例分析、评估,验证了所提模型和算法的有效性,并通过增加样本数据点获得了更加精确的预测结果。The meteorological factors that affect photovoltaic （PV） system output power are analyzed, and PV output power forecasting model based on weather type clustering and LS-SVM is proposed by combing PV system actual operation data and weather information. Solar irradiation time, temperature, relative humidity and so on are selected as the meteorological feature vector. Through the calculating of weighted Euclid distance of each vector, the best clustering set is selected and the training samples are determined to better reflect actual weather information of the day to be predicted. Meteorological characteristics, corresponding PV output power of the best clustering day and meteorological characteristics of the day to be predicted are taken as the input of the trained LS-SVM mod- el, and the PV output power of the corresponding time of the day to be predicted is used as the output. Finally, the effectiveness of the model and algorithm is verified by analyzing and evaluating the actual examples. More accurate forecasting results are obtained by increasing the sampling points.

关 键 词：聚类 最小二乘支持向量机 

分 类 号：TM615[电气工程—电力系统及自动化]