Stacked ensemble learning-based approach for a coordinated optimal energy management framework in renewable energy assisted microgrid systems

The paper proposes an optimal energy management framework for renewable energy–assisted microgrid (MG) systems using a machine-learning-based approach. Smart and intelligent real-time energy management techniques enable flexible and efficient utilization of distributed energy resources (DERs); however, accurate prediction of renewable generation and variable loads remains challenging. To address this issue, a stacked ensemble learning (SEL) model is developed to achieve more reliable and accurate renewable energy forecasting. One year of solar irradiance data collected for Roorkee, Uttarakhand, India (29.8543 o N , 77.8880 o E ) is used for model training and validation. The SEL framework integrates multiple individual learners, exploiting their complementary strengths and diverse data interpretations to enhance overall prediction accuracy. This improved forecasting capability supports coordinated operational optimization of both grid-connected and off-grid (isolated) renewable-assisted MG systems. The Performance evaluation demonstrates that the proposed SEL model outperforms conventional methods such as GBM, XRT, DRF, etc. The achieved RMSE, MSE, MAE, RMSLE, and mean residual deviance values are 0.4072, 0.1658, 0.1829, 0.0968, and 0.1658, respectively, representing the lowest errors among all compared models. Based on accurate forecasts, optimal day-ahead dispatch strategies are derived for grid-connected and off-grid MG configurations that satisfy operational constraints. The AutoML technique is employed to automate the application of the proposed algorithm, and its performance is compared with baseline models such as ARIMA and Prophet. The effectiveness of the proposed strategy is validated using a hardware-in-loop (HIL) setup.