Federal learning-based power data safe and efficient sharing method

Abstract

The invention belongs to the field of implementation methods of intelligent power data sharing, and particularly relates to a safe and efficient power data sharing method based on federal learning. Comprising the following steps: S100, deploying data acquisition equipment in each electric power company; s200, storing the interaction data obtained by each electric power node in a local database of each electric power company; s300, the coordinator sends the initialized model parameter w to each power node; s400, each power node performs model training on stored data through an initial model parameter w, and S500, each power node performs homomorphic encryption and blinding operation on the obtained value of the convex loss function and the model parameter w, and transmits the value and the model parameter w to a coordinator; s600, placing and summarizing by the coordinator to obtain a new model parameter w; and S700, the coordinator judges whether the current new model parameter w is converged or not according to the calculation result. The privacy data of each electric power company is ensured, and sensitive data leakage is effectively prevented.

Description

technical field [0001] The invention belongs to the field of realization methods for intelligent power data sharing, and in particular relates to a safe and efficient power data sharing method based on federated learning. Background technique [0002] With the rapid development of the Internet of Things technology, more and more industrial systems are connected to it. At the same time, due to the arrival of Industry 4.0, related technologies in the field of artificial intelligence have also been introduced into the industrial field. More and more industrial data in the machine learning model Among them, the field of electric energy is more widely used, but at the same time, due to the limited amount of relevant data of each power company node and the need for privacy protection, there is a contradiction with the wide application of machine learning technology, so for The industry often introduces federated learning technology to deal with the above phenomena. [0003] As a ...

AI Technical Summary

Problems solved by technology

Based on this benchmark algorithm, Bonawitz K et al. constructed a scalable federated learning algorithm (Bonawitz K , Eichner H , Grieskamp W , et al. Towards Federated Learning at Scale: SystemDesign[J].2019.), which can Effectively solve the problem of unreliable terminal connection and sudden interruption of equipment. However, the algorithm has slow convergence speed and low parallel efficiency. At the same time, it fails to solve the problem of non-independent and identical distribution of data, and is not suitable for the field of electric energy

Xie et al proposed an asynchronous federated learning algorithm (Xie C , Koyejo S , Gupta I .Asynchronous Federated Optimization[J]. 2019.), and proposed a time-delay model weight update method, but if the algorithm If the node upload delay time is too long, it will have a great impact on the final result, and cannot meet the needs of safe and efficient data sharing in the power field

Based on this, a federated learning method based on node clustering is proposed to solve the problem that node data cannot be shared safely and efficiently due to non-independent and identical distribution in the power field.

Images