Power industry lifting operation peccancy detection method based on reinforced federal learning

Abstract

The invention discloses a power industry hoisting operation peccancy detection method based on reinforced federated learning, and the method comprises the following steps: S1, employing federated learning C, training a node A and a node B through employing local data, and obtaining a model; S2, inputting the model obtained in the step S1 into a reinforcement learning module, performing model fusion by using reinforcement learning DQN, and adjusting weights of the A node model and the B node model; S3, enabling the reinforcement learning module to generate a reinforcement fusion model through reinforcement learning; s4, performing model fusion on the models of the node A and the node B by the central node of the federated learning C by using a reinforced fusion model and weighted average; S5, issuing the fused model to the node A and the node B; and S6, repeating the steps S1 to S5 until model training is completed. The method has the beneficial effects that the joint modeling effect of federal learning is ensured by using reinforcement learning, high-quality nodes are selected to jointly establish a model, and the influence of a heterogeneity problem is reduced.

Description

technical field [0001] The invention relates to the technical field of detection methods for lifting operation violations in the electric power industry, in particular to a detection method for lifting operation violations in the electric power industry based on enhanced federated learning. Background technique [0002] As an emerging technology, federated learning's ability to protect the private data of individuals and organizations coincides with the concept of technology in the new infrastructure to promote public value. At the same time, it helps research institutions and organizations in various industries to release data value in compliance. received widespread attention. The obvious benefit of federated learning is distributing the quality of knowledge across a large number of devices without centralizing the data used to optimize and train the model, while this approach also enables improving the quality of centralized machine learning models while maintaining the p...

AI Technical Summary

Problems solved by technology

[0006] (2) Weighted fusion: Weighted fusion means that the weight of each node is set according to the amount of data in each byte and the performance of the device during model fusion. This will have a better effect than the average fusion model, but the disadvantage is that it cannot be quantitatively evaluated. For each node, which nodes to integrate rely on experience

Images