A power system typical operating mode extraction method based on deep time series aggregation

Abstract

The application discloses a kind of based on depth time series aggregation power system typical operating mode extraction method, the method includes the following steps: data pre-processing, obtain time series dataset;Gram angle and field image coding are carried out to time series dataset, generate image sequence;Double-path feature extraction network is constructed, and stable feature and regional association feature are extracted to image sequence;Dimensionality reduction is carried out to feature and is spliced, and operating mode representation is obtained;Based on Gaussian mixture model and operating mode representation, construct depth clustering model, and optimize depth clustering model by end-to-end joint training;Extract clustering center as typical operating mode;The stability evaluation, key association feature and dispatching guidance suggestion of each typical operating mode are output, and form typical operating mode knowledge base.The application solves the problems of insufficient dynamic evolution information description, one-sided feature expression and weak physical meaning in traditional power system operating mode extraction, and insufficient analysis efficiency and adaptability.

Images