基于层次聚类分场景的光伏汇聚趋势量化方法

PHOTOVOLTAIC CONVERGENCE TREND QUANTIFICATION METHOD BASED ON HIERARCHICAL CLUSTERING AND SCENARIOS

持续功率曲线能反映长时间波动特性规律,通过研究已建设光伏集群持续功率曲线,建立预测模型揭示不同规模集群的汇聚演化规律,最终得到待建光伏集群的持续功率曲线。首先,利用层次聚类算法确定光伏集群汇聚规模的分层顺序,得到装机容量逐层递增的光伏集群,并提出汇聚效应指标验证顺序的有效性;其次,为了更好地判断和预测光伏持续功率曲线的变化趋势,对持续功率曲线进行出力场景划分;最后,为避免单一模型预测偏差,在各出力场景下,通过改进的信息熵组合预测模型掌握汇聚过程中规模演变规律,完成规划待建设集群持续功率曲线的预测。利用河北某地区实测数据仿真结果表明:验证聚类方法得到的集群分层顺序更能体现汇聚效应,并有效提高预测精度;出力场景划分准确刻画集群持续功率曲线汇聚趋势;通过模型对比表明分场景下改进信息熵组合预测模型更能精确完成待建光伏集群持续功率特性的量化分析。

The continuous power curve can reflect the law of long-term fluctuation characteristics.By studying the known continuous power curve of photovoltaic clusters,a prediction model is established to reveal the convergence evolution law of clusters of different scales,and finally,the continuous power curve of the photovoltaic cluster to be built is obtained.Firstly,the hierarchical clustering algorithm is used to determine the hierarchical order of the aggregation scale of photovoltaic clusters,and the photovoltaic clusters with the installed capacity increasing layer by layer are obtained,and propose aggregation effect indicators to verify the effectiveness of the sequence.Secondly,in order to better predict the change trend of the photovoltaic continuous power curve,and divide the output scene of the continuous power curve.Finally,in order to avoid the prediction deviation of a single model,in each output scene,the improved information entropy combination prediction model is used to grasp the scale evolution law in the aggregation process and complete the prediction of the continuous power curve of the cluster to be built.The simulation results using the measured data in a certain area in Hebei show that the cluster hierarchical order obtained by verifying the clustering method can better reflect the convergence effect and effectively improve the prediction accuracy;the output scene division accurately describes the convergence trend of the continuous power curve of the cluster;and the improved information entropy combination prediction model can more accurately complete the quantitative analysis of the continuous power characteristics of the photovoltaic cluster to be built.