Regression based anomaly detection in electric vehicle state of charge fluctuations through analysis of electric vehicle charging infrastructure Data

With the increase in the number of electric vehicles (EV), there is a need for the development of the EV charging infrastructure (EVCI) to facilitate fast charging, thereby mitigating the EV congestion at charging stations. The role of the public charging station depot is to charge the vehicle, prioritizing the achievement of the desired state of charge (SoC) value for the EV battery or charging till the departure of the EV, whichever occurs first. The integration of cyber and physical components within EVCI defines it as a cyber physical power system (CPPS), increasing its vulnerability to diverse cyber attacks. When an EV interfaces with the EVCI, mutual exchange of data takes place via various communication protocols like the Open Charge Point Protocol (OCPP), and IEC 61850. Unauthorized access to this data by intruders leads to cyber attacks, potentially resulting in consequences like energy theft, and revenue loss. These scenarios may cause the EVCI to incur higher charges than the actual energy consumed or the EV owners to remit payments that do not correspond adequately to the amount of energy they have consumed. This article proposes an EVCI architecture connected to the utility grid and uses the EVCI data to identify the anomalies or outliers present in the EV transmitted data, particularly focusing on SoC irregularities. The proposed methodology involves utilizing a ridge regression based machine learning (ML) model for predicting changes in the SoC. The adversaries have the capability of spoofing these change in SoC values, consequently making the EVCI incapable of achieving the desired task. Three distinct spoofing techniques namely, decimal shifting, incremental array spoofing, and random spoofing are implemented on the data and subsequently tested with the proposed methodology. The results show that the proposed methodology detects the anomaly accurately and also classifies the type of spoofing that causes the anomaly.