Analysis Model of Non-Destructive Regional User Power Consumption Behavior Based on Federated Learning

In response to the national "dual carbon" policy, the power supply strategy of power enterprises should be changed to the direction of "personalization" of regional users, and the analysis of user electricity consumption behavior has become an important task in the reform of the power system, and the existing analysis of regional user electricity consumption behavior cannot meet the requirements of privacy and intelligence. In order to solve this problem, this paper proposes a lossless regional user power consumption behavior analysis model based on federated learning, introduces the TPA-LSTM model to complete the user power consumption behavior analysis and prediction task, protects the privacy of user data, reduces the risk of data circulation by building a data sharing mechanism between users and model collaborators, designs a model lossless aggregation scheme, and realizes lossless aggregation of model parameters.