晴天光伏发电功率的日内变化规律及预测方法研究

Daily Variation Laws and Prediction Methods in Photovoltaic Power Generation on Sunny Days

针对晴天天气类型,研究了光伏发电功率的日内和年内尺度的变化规律,根据晴天出力变化速率曲线的特性,提出晴天光伏发电功率预测模型,在晴天光伏发电功率预测模型的基础上,根据光功率实时监测信息,加入分阶段实时修正系统,实现了日前预测与超短期预测相融合,提高了预测精度.黄河上游共和光伏电站实例应用结果表明:所提出的晴天光伏发电功率预测模型的总体平均预测误差在2%以下,用于光伏发电功率预测的其他方法如线性回归、神经网络、支持向量机等在相同条件下的预测误差区间为8%~20%,故所提出的方法预测精度有大幅度提高.该模型的输入值都是基于光伏电站自身的出力特性提取的,不需要辐射量、云层、气压、降水量等难以获取的信息,是一种简单并实用的创新预测方法.

This paper studies the variation law of the daily and annual scale photovoltaic(PV)power generation according to"sunny weather mode".A sunny photovoltaic power prediction model is proposed,and the phased realtime correction system based on optical power monitoring information is added.The method realizes the power generation predictability of photovoltaic power plants that combines previous predictions with ultra-short-term prediction.Application results of the Republican Photovoltaic Power Station in the upper reaches of the Yellow River indicate that the overall average prediction error of the proposed sunny photovoltaic power prediction model is≤2%.Other methods used for PV power prediction such as linear regression,neural networks,and support vector machines have error detection intervals of 8%-20%under the same conditions.Therefore,the prediction accuracy of the proposed method is greatly improved.The proposed model does not require information such as the amount of radiation,cloud layer,air pressure,and precipitation,which are difficult to obtain.The model provides a simple,practical,and innovative prediction method.