A Time Space Feature Extraction Method for Scene Analysis Based on Spectral Clustering Theory

A method for extracting time space feature volume of power system scenes based on spectral clustering is proposed. Through downscaling and correlation processing of massive high-dimensional data, massive variables with both temporal and spatial characteristics dispersed in the power grid are clustered according to the similarity of their influence on the operation mode, and the time space feature volume applicable to the analysis of the power grid scenario is extracted. Classification and screening based on the extracted feature quantities, generating scenario patterns classified in time and space scales, and predicting the short-term operation mode of the power grid by analyzing historical electricity consumption data and multiple influencing factors