Urban Integrated Energy Edge Collaboration and Privacy Protection Based on the Federated Learning Framework

Abstract

In the context of building a low-carbon integrated energy city, the energy data of electricity, water, heat, gas and other in city belong to different public utility management departments, and there are information barriers and data privacy protection issues between different energy systems. Under the premise of ensuring data security, aiming at the problem of training algorithm models for different energy data, the energy edge collaboration architecture based on homomorphic encrypted federated learning is proposed firstly, and then the urban comprehensive energy federated learning model is established to realize the local training model and transmit the updated model parameters to the server. Finally, the prediction of multivariate short-term loads by federated learning framework training and centralized training neural network models is verified by comparative experiments, and the results prove the feasibility and superiority of urban integrated energy edge collaboration and privacy protection based on federated learning framework in the field of urban integrated energy.