Electric power inspection image anomaly detection method and system based on federal learning

Abstract

The invention discloses an electric power inspection image anomaly detection method and system based on federal learning. The method comprises the following steps: adopting a pre-trained anomaly detection model as a server initial model, and distributing the model to each participant; participants perform training based on a local electric power inspection image data set and upload cut gradient information to a server; the server judges the data quality on the auxiliary model by using a verification set according to the gradients uploaded by the participants, and rejects the participants with poor data quality based on a set threshold value; performing weighted aggregation on the model gradients based on data volume imbalance and data quality indexes to obtain global gradient parameters and update an anomaly detection global model, and then distributing a new model to each participant; and circulating the steps until the global model converges. According to the method provided by the embodiment of the invention, the accuracy of the local electric power inspection image anomaly detection model of each participant can be improved.

Description

technical field [0001] The invention belongs to the field of electric equipment detection, and in particular relates to a federated learning-based method for abnormal detection of electric power inspection images. Background technique [0002] In order to be able to detect abnormalities of power equipment in time so as to effectively eliminate potential safety hazards, the power operation and maintenance department needs to conduct regular inspections of power equipment. However, manual inspection is time-consuming, labor-intensive and inefficient, and the accuracy of anomaly detection depends on the professional level of inspectors. With the rapid development of artificial intelligence technology, abnormal detection of power inspection images based on deep learning has gradually become a reality. [0003] However, in the actual industrial production environment, the abnormal detection of power inspection images will face the following problems: 1. The power inspection imag...

AI Technical Summary

Problems solved by technology

[0003] However, in the actual industrial production environment, the abnormal detection of power inspection images will face the following problems: 1. The power inspection image data exists in various power companies in a fragmented manner, forming a "data island" with a small amount of local data. It is not enough to train a better network, and these scattered data are non-independent and identically distributed, so it is not feasible and costly to gather them together for training; 2. According to the data security regulations of the State Grid, it is prohibited Electronic collection business data and devices are provided to third parties in the society

Images