计及时空分布特性的光-荷典型场景提取方法

A Method for Extracting Typical Scenes of Photovoltaic and Load Considering Temporal and Spatial Distribution Characteristics

针对含光伏配电网中典型场景生成对光伏、负荷的时空相关性考虑不充足的问题,提出基于拉普拉斯矩阵降维及高斯混合模型的光伏、负荷典型场景提取法。首先利用光伏、负荷的历史运行数据构建光伏、负荷的日场景描述矩阵;再通过拉普拉斯特征映射的降维,以及高斯混合模型得到描述矩阵的概率密度函数;最后经过Wasserstein距离筛选,完成光伏、负荷典型日场景提取。基于广东某地含光伏配电网的运行数据进行实验,对基于k-means聚类的场景生成法、拉丁超立方采样法以及所提场景联合方法进行对比,结果显示所提方法生成的典型场景与配电网实际运行场景的误差更小,速度更快,数量更少。

In view of the problem of insufficient consideration of the spatiotemporal correlation between photovoltaic(PV)and load in typical scenarios generated in the distribution networks with PV,this paper proposes a method for extracting typical PV and load scenarios based on Laplace matrix dimensionality reduction and Gaussian mixture model(GMM).Firstly,it uses the historical operating data of PV and load to construct a daily scenario description matrix for PV and load.Then,the probability density function of the description matrix is obtained through dimensionality reduction of Laplacian eigenmaps(LE)and GMM.Finally,after filtering through Wasserstein distance,the extraction of typical solar and load daily scenarios is completed.The operation data from a PV distribution network in Guangdong province is uses for the experiment.The paper compares the scenario generation method based on k-means clustering,Latin hypercube sampling method and the proposed scenario joint method.The results show that the typical scenarios generated by the proposed method have smaller errors,faster speed,and fewer quantities compared to the actual operating scenarios of the distribution network.