Adaptive Learning‐Based Hybrid Prediction Energy Management Scheme for Sustainable EV Charging in Standalone DC Microgrids

The integration of electric vehicle charging stations (EVCSs) with standalone microgrids (MGs) offers a sustainable alternative to grid‐dependent power supply. However, managing stochastic EV charging demand while considering intermittent renewable power‐flow dynamics and limited battery storage system (BSS) capacity in isolated MG configurations poses significant challenges. To address these issues, an adaptive learning‐based hybrid prediction energy management scheme (AL‐PEMS) is proposed. Unlike traditional regression‐based predictive energy management schemes, AL‐PEMS employs a novel hybrid prediction framework that integrates a Polynomial Regression Predictive Model with a hierarchical adaptive window‐based learning algorithm to accurately predict stochastic EV charging demand and dynamically regulate dispatchable power flow. The adaptive learning‐assisted predictive energy manager hierarchically refines successive prediction windows according to the severity of BSS‐aware prediction errors, thereby reducing excessive BSS loading. The effectiveness of the proposed scheme is validated using real‐world EV charging data from 32,920 charging events at Oslo charging stations, while its real‐time feasibility is demonstrated through controller hardware‐in‐the‐loop testing. A comparative performance analysis of the proposed AL‐PEMS against rule‐based adaptive power management strategies, regression‐based model predictive control energy management schemes, and an adaptive model predictive control scheme demonstrates significant improvements, including 19.82%, 8.43%, and 7.87% enhancement in EVCS reliability, 17.60%, 20.20%, and 18.13% improvement in sustainability, 35.23%, 28.36%, and 28.36% reduction in BSS loading, and 14.16%, 9.20%, and 8.89% improvement in DC MG voltage regulation under stochastic EV charging conditions.