EV charging station monitoring framework—A case study of Croatia: Performance analysis and anomaly detection under pandemic situation

Abstract

This paper investigates the pandemic impact on electric vehicle (EV) monitoring framework designed for analysing and tracking of EV charging data in the Croatian electric vehicle charging station grid. The developed framework encompasses multiple features of the EV charging grid behaviour, notably: (i) global predictions of the utilized energy, (ii) number of charging vehicles, and (iii) total time spent while charging. In this paper, we analyse framework performance and key performance indicator (KPI) values during the pre-lockdown period, soft-lockdown period, and post-lockdown lift-off period to show how the pandemic situations and governmental lockdown regulations can impact EV grid performance and EV charging business. In addition to that, we test a daily framework anomaly detection flow that tracks the global predictive performance of the predictive models on the EV grid, to detect if the framework detects this to be a situation that is novel (anomalous) during the soft-lockdown period. Results show that changes in the behaviour during the lockdown period reduce electric demand on EV charging spots, with customers having lower demand and similar behaviour as in other European countries. Furthermore, the soft-lockdown period is detected as a period with more extreme anomalous values for supervised models, and this is shown as an increase in residual anomaly values, flagging 3 times more days as an extreme anomaly than in the normal periods.