光伏发电功率的智能预测算法  

Intelligent Forecasting Algorithm for Photovoltaic Power Generation

作　　者：程泽[1] 韩丽洁[1] 李思宇[1] 巩力[1] 

机构地区：[1]天津大学电气与自动化工程学院,天津市300072

出　　处：《电力建设》2014年第7期34-39,共6页Electric Power Construction

基　　金：国家自然科学基金项目(61374122)

摘　　要：光伏发电系统的出力具有强烈的波动性,为了减轻其对电网的冲击,有必要进行光伏出力预测。提出了一种基于灰色关联度分析(gray relational analysis,GRA)和最小二乘支持向量机(least square support vector machine,LSSVM)方法对光伏出力进行预测,该方法是传统直接预测和间接预测方法的结合,分析了辐照度、天气类型等对光伏输出功率的影响。通过GRA选择训练样本,使样本更全面地反映预测日的天气属性;然后运用LSSVM提前24 h预测输出功率,利用天津市太阳能光电建筑示范项目的实测数据对该预测模型进行了测试与评估,算例结果表明,所提出的GRA-LSSVM的预测方法具有较高的预测精度。Photovoltaic （PV） power forecast is significant to reduce the impact of PV generation on the power system since the output of PV system has a strong randomness. This paper presented a method based on least square support vector machine （LSSVM） and gray relational analysis （GRA） to predict the PV output power, which was a combination of traditional forecasting methods including direct and indirect forecasting. First, it analyzed the effects of irradiance and weather types on PV output, and selected training samples through GRA, so that the training samples could reflect the weather attributes at forecast days more exactly. Then the predictive model could forecast output power 24 hours ahead with using LSSVM, and it had been tested and evaluated by the measured data of solar - PV building demonstration project in Tianjin. The results show that the prediction method of GRA - LSSVM has better orediction accuracy.

关 键 词：灰色关联度分析(GRA) 最小二乘支持向量机(LSSVM) 光伏发电 功率预测 

分 类 号：TM615[电气工程—电力系统及自动化]