Aircraft lead-acid storage battery anomaly detection semi-supervised method based on isolation

Abstract

The invention discloses an aircraft lead-acid storage battery anomaly detection semi-supervising method based on isolation. The method comprises the steps that data such as voltage and resistance of an aircraft lead-acid storage battery are acquired, and resampling is performed to reduce redundancy; t-SNE dimension reduction is carried out on the processed data to obtain two dimensions, visualization is carried out, and batteries of different clusters are divided; battery data of different clusters arenormalized to be between 0 and 1 to the maximum extent and the minimum extent, data with labels are taken as input of an anomaly detection algorithm iForest or iNNE based on isolation, calculation is carried out to obtain an abnormal value of each group of batteries, and parameters are adjusted to maximize an effect measurement standard AUC value; after the obtained abnormal values are sorted, the corresponding abnormal value when the sensitivity reaches 0.9 or above is taken as a threshold value of division abnormality; and according to the determined optimal algorithm parameter and threshold corresponding to a certain feature of a certain cluster, an abnormal value of the remainingunlabeled data is calculated, if the abnormal value is higher than the threshold, the battery is determined to be an abnormal battery, and otherwise, the battery is determined to be a normal battery.

Description

technical field [0001] The invention relates to an isolation-based semi-supervisory method for abnormal detection of aircraft lead-acid batteries, and belongs to the technical field of abnormal detection and data mining. Background technique [0002] In recent years, lead-acid batteries have been successfully used in various vehicles such as battery cars, automobiles, forklifts, and gradually expanded to military communications, navigation, aerospace and other fields. As the number of charge and discharge cycles increases, the effective capacity of the lead-acid battery will gradually decrease. Lead-acid batteries can fail when the battery capacity falls below a threshold. With the rapid development of the aviation industry and the increasingly busy aviation tasks, the requirements for the safety and reliability of the power system are also getting higher and higher. Lead-acid batteries play a role that cannot be ignored as backup power. However, at present, the cost of u...

AI Technical Summary

Problems solved by technology

However, these algorithms still have some shortcomings, such as the latent cost of calculating the nearest neighbor and the high sensitivity of the neighborhood size selection

In addition, due to the characteristics of unsupervised algorithms, that is, they completely get rid of the dependence on data labels, and there is no partial label to measure their detection effect, so these anomaly detection algorithms often have the problem of optimal parameter adjustment.

Images