Feature extraction and source–load collaborative analysis method for distribution network

The data from new distribution system is increasingly multi-dimensional and massive. When the distribution network participates in virtual grid division, the deep mining of distribution network data is required. In this study, a feature extraction and source–load collaborative analysis method for distribution network is proposed to master the suitability between renewable energy generation and various loads. Firstly, this method realizes raw data preprocessing through anomaly identification and reconstruction. It then uses information entropy to improve traditional piecewise aggregation approximation to reduce data dimension while considering the fluctuations of data. Finally, the method extracts the features of source–load data through spectral clustering and obtains quantitative analysis results of the matching degree between source and load through collaborative analysis. The effectiveness of the proposed method is verified by load power data and photovoltaic output data.