Regional electricity consumption medium and long term prediction method based on federated learning

Abstract

The invention discloses a medium and long term prediction method for regional electricity consumption based on federal learning, and belongs to the technical field of electrical engineering. According to the region electricity consumption medium and long term prediction method based on federated learning, federated training is carried out on electricity consumption prediction models of multiple regions at the same time, on the premise that data privacy is protected, training samples are enriched, and the generalization ability of the models is effectively improved; a local basic model of federal learning adopts an LSTM network, and on the basis of traditional influence factors of electricity consumption of economy, society and the like, carbon emission is added to input characteristics of the model, so that an electricity consumption prediction result is more in line with a future low-carbon development path. The method is a data driving method, does not involve any explicit modeling, has better adaptive capacity for a high-dimensional nonlinear problem compared with a traditional statistical method and a regression model, can mine the potential change rule of the electricity consumption through self-learning, and improves the prediction accuracy.

Description

technical field [0001] The invention relates to the technical field of electrical engineering, in particular to a medium- and long-term forecasting method for regional electricity consumption based on federated learning. Background technique [0002] Under the background of the "carbon peak, carbon neutral" ("double carbon") strategy, the future direction of power system transformation is to build a new power system with new energy as the main body. Accurate mid- and long-term forecasts of regional electricity consumption can guide regional new energy, traditional thermal power installed capacity and other power planning, and help the government to formulate "dual carbon" strategic goals. Existing machine learning and deep learning methods have achieved some results in medium and long-term forecasting of regional electricity consumption, but in actual electricity consumption forecasting, the accurate electricity consumption data recorded in various regions are only years or ...

AI Technical Summary

Problems solved by technology

[0003] In addition, the existing regional electricity consumption forecasting methods only consider economic and social factors, ignoring the impact of carbon emissions on electricity consumption under the background of "double carbon", and the prediction results may not be in line with the future "double carbon" Target

Images