<p>The increasing integration of renewable energy sources into power systems has accelerated the deployment and usage of microgrids. However, the stochastic nature of renewable energy generation and fluctuating power demand pose significant challenges to maintaining grid stability. To address these challenges, this paper proposes an intelligent Model Predictive Control (MPC) framework for optimal power flow management in microgrids, with the objective of enhancing operational resilience, reducing diesel fuel consumption, and preventing blackouts through coordinated electric vehicle (EV) charging and discharging. The proposed MPC is formulated as a nonlinear optimization problem that minimizes operational costs while satisfying dynamic microgrid constraints. A Passive MPC strategy, which relies on predefined EV availability schedules without predictive forecasting, is first investigated to assess the impact of EV participation on microgrid performance. Simulation results show that allowing EV support outside critical hours (9 AM to 7 PM) achieves fuel and CO<InlineEquation ID="IEq1"> <EquationSource Format="TEX">$_2$</EquationSource> <EquationSource Format="MATHML"><math> <mmultiscripts> <mrow /> <mn>2</mn> <mrow /> </mmultiscripts> </math></EquationSource> </InlineEquation> emission reductions of approximately 9.11% compared to a baseline scenario without EV integration. When EVs are available throughout the entire day, fuel consumption and CO<InlineEquation ID="IEq2"> <EquationSource Format="TEX">$_2$</EquationSource> <EquationSource Format="MATHML"><math> <mmultiscripts> <mrow /> <mn>2</mn> <mrow /> </mmultiscripts> </math></EquationSource> </InlineEquation> emissions are further reduced by up to 33.39%, demonstrating the significant potential of vehicle-to-grid participation. Building upon these results, a Deep Learning–based MPC (DL-MPC) framework is developed to enable proactive and informed control decisions. An LSTM-based forecasting model is employed to predict day-ahead load demand, wind power generation, and solar irradiance, allowing the MPC to anticipate system conditions and mitigate blackout risks more effectively. The forecasting models exhibit strong predictive accuracy, with RMSE values of 3.9194 for wind power, 0.0908 for load demand, and 107.8 for solar irradiance, and corresponding MAPE values of 18.64%, 22.00%, and 22.3%. The proposed framework highlights the potential of combining predictive control with vehicle-to-grid strategies to enhance the sustainability, and efficiency of microgrid operations while effectively mitigating the risk of blackouts.</p>

A deep learning based predictive control method for enhancing microgrid resilience

期刊论文
发表日期：

2026年8月5日

查看全文

Hussein A. Taha,
Ahmed Abdelrahman,
Abdelhamid Mammeri

摘要

The increasing integration of renewable energy sources into power systems has accelerated the deployment and usage of microgrids. However, the stochastic nature of renewable energy generation and fluctuating power demand pose significant challenges to maintaining grid stability. To address these challenges, this paper proposes an intelligent Model Predictive Control (MPC) framework for optimal power flow management in microgrids, with the objective of enhancing operational resilience, reducing diesel fuel consumption, and preventing blackouts through coordinated electric vehicle (EV) charging and discharging. The proposed MPC is formulated as a nonlinear optimization problem that minimizes operational costs while satisfying dynamic microgrid constraints. A Passive MPC strategy, which relies on predefined EV availability schedules without predictive forecasting, is first investigated to assess the impact of EV participation on microgrid performance. Simulation results show that allowing EV support outside critical hours (9 AM to 7 PM) achieves fuel and CO $_2$ $_{2}$ emission reductions of approximately 9.11% compared to a baseline scenario without EV integration. When EVs are available throughout the entire day, fuel consumption and CO $_2$ $_{2}$ emissions are further reduced by up to 33.39%, demonstrating the significant potential of vehicle-to-grid participation. Building upon these results, a Deep Learning–based MPC (DL-MPC) framework is developed to enable proactive and informed control decisions. An LSTM-based forecasting model is employed to predict day-ahead load demand, wind power generation, and solar irradiance, allowing the MPC to anticipate system conditions and mitigate blackout risks more effectively. The forecasting models exhibit strong predictive accuracy, with RMSE values of 3.9194 for wind power, 0.0908 for load demand, and 107.8 for solar irradiance, and corresponding MAPE values of 18.64%, 22.00%, and 22.3%. The proposed framework highlights the potential of combining predictive control with vehicle-to-grid strategies to enhance the sustainability, and efficiency of microgrid operations while effectively mitigating the risk of blackouts.