Identification of Charging Behavior for Electric Bicycles based on Supervised Fisher Classifier

The abnormal charging behavior of electric bicycles may give rise to serious fire accidents. In order to avoid these risks, it is important to monitor the charging process of electric bicycles. Non-intrusive load monitoring (NILM) technology is proven to be an effective tool in this field. Existing studies mainly focus on using large amounts of transient data to learn the charging characteristics of electric bicycles, while transient data is not easily accessible considering smart meters installed at residential customer’s side mainly provide steady-state data for analysis. In light of this issue, we proposed a non-intrusive identification method of charging behavior for electric bicycles based on supervised fisher classifier. Firstly, the collected electrical data is preprocessed by filling in the missing values and deleting abnormal data. Afterwards, the supervised fisher classifier is utilized to classify the data and identify the charging behavior of electric bicycles. Simulation is carried out based on realistic data in MATLAB. Results show that the proposed method can accurately identify the charging behavior of electric bicycles.