Real-time estimation of aggregated electric vehicle charging load based on representative meter data

The uncontrolled integration of numerous electric vehicles (EVs) brings great uncertainty to grid regulation. Real-time monitoring of widely dispersed EV charging load meter data requires a large number of efficient data acquisition equipment and transmission channels, which brings high investment and operating costs. To address this challenge, this paper proposes a data-driven method for real-time estimation of aggregated EV charging load. A maximum relevance minimum redundancy selection method based on pearson correlation coefficient (mRMR-P) is proposed to select a representative subset of EV charging station (EVCS) meter data and eliminate redundancy. Subsequently, a deep learning model constructed in this paper extracts the load features and temporal relationships from the selected representative meter data to achieve aggregated estimation of EV charging load. Additionally, to address the issue of model degradation due to changes in EV users' charging behavior over time, an adaptive window concept drift detection (CDD) method based on the model's input-output mapping relationship is proposed. Finally, the proposed method is validated using real data from residential and public EVCS in Hangzhou, China. Experimental results demonstrate the effectiveness and superiority of the proposed method.