考虑遮挡因素的分布式屋顶光伏日前母线功率预测

Prediction for Day-ahead Power of Distributed Rooftop Photovoltaic Considering Occlusion Factor

针对分布式屋顶光伏装机受到遮挡问题导致分布式光伏用户的出力特性存在差异的情况,提出考虑遮挡因素的变电站母线级分布式屋顶光伏功率日前预测模型。首先根据变电站的经纬度计算出每日光伏出力的起始时间和截止时间;其次根据每个光伏用户的历史功率数据分析周围建筑物对光伏遮挡造成的影响,采用形状距离作为度量指标进行凝聚层次聚类;然后利用辐照度数据对每一类分布式光伏用户建立BP功率预测模型,将功率预测值通过LSTM神经网络修正得到最终预测值;最后将每一类分布式屋顶预测结果相加,获得变电站母线级分布式屋顶光伏日前功率预测值。实例分析表明,所提出的预测方法精度高,与不考虑遮挡因素的预测方法相比,均方根误差显著降低。

In view of the difference of output characteristics of distributed PV users due to the occlusion problem of distributed rooftop photovoltaic(PV)installation,a day-ahead prediction model of distributed rooftop PV power with substation bus level considering the occlusion factor is proposed.Firstly,the start time and cut-off time of daily PV output are calculated according to the latitude and longitude of the substation.Secondly,according to the historical power data of each PV user,the influence of the surrounding buildings on the photovoltaic occlusion is analyzed,and the shape distance is used as the metric to carry out the cohesive hierarchical clustering.Then,the irradiance data is used to establish a Back Propagation(BP)power prediction model for each type of distributed photovoltaic user,and the power prediction value is corrected by the Long Short-Term Memory(LSTM)neural network to obtain the final predicted value.Finally,the prediction results of each type of distributed roof are added to obtain the day-ahead power prediction of distributed rooftop PV with the substation bus level.Using the actual photovoltaic power station for case analysis,the results show that high accuracy of the prediction method proposed in this paper reaches,and the Root Mean Squared Error(RMSE)is significantly reduced compared with the prediction method without considering the occlusion factor.