A Novelty Detection Method for Non-Intrusive Electric Bicycle Chargers

In this paper, we propose a novel non-invasive elec-tric bicycle charger detection method that aims to identify electric bicycle chargers from the perspective of their potential electrical safety threats among traditional household appliances. We consider electric bicycle chargers as anomalous loads in household appliances with an eye to solving the problem of dichotomizing electric bicycle chargers from other appliances, and propose easi-ly distinguishable V-I trajectory features through feature screening. In this paper, we adopt three different feature screening methods, comprehensively consider the relationship between 26 features in different aspects, and finally extract the optimal sub-set of features in three different dimensions. Our approach applies the anomaly detection algorithm OCSVM and trains these three optimal feature subsets under an integrated learning framework, which significantly improves the testing accuracy of electric bicycle chargers. This research provides an innovative and effective method for identifying potential safety issues in elec-tric bicycle chargers, which is expected to playa key role in improving the overall safety of the charging process.