Machine learning-enhanced coordination of home-microgrids for resource compensation in large-scale energy systems

To achieve the goals of sustainable zero-carbon cities, the replacement of fossil fuel power plants with clean alternatives has long been seen as a challenge. Home microgrids (HMGs), integrated with renewable energy sources (RESs), energy storage system (ESS), and electric vehicle (EV), offer a cost-effective pathway but are constrained by individual capacities. In order to solve this, a smart neighborhood (SN) is made up of interconnected HMGs that can function as a cloud ESS for the main grid. The energy transition depends on these interrelated HMGs being managed effectively. However, traditional control approaches are insufficient in SNs due to the complexity of managing various resources, variables, and restrictions. This study addresses the complexity of managing interconnected HMGs by proposing an artificial intelligence-based energy management system (AI-EMS) using a deep reinforcement learning (DRL) framework based on the deep deterministic policy gradient (DDPG) algorithm. The AI-EMS ensures optimal utilization of HMGs local sources, minimizes energy costs, and supports the main grid by providing ancillary services. By treating the SN as a unified cloud ESS, the proposed approach reduces reliance on fossil fuels, enables efficient bi-directional energy exchange with the main grid, and enhances energy flexibility and resilience. The proposed AI-EMS approach is validated through MATLAB/Simulink simulations under various case studies (CSs). According to simulation results, compared to the baseline scenario (CS1), energy imported from the main grid decreased by 21% (from 533.4 kWh to 421.2 kWh) under CS2 and increased by 7.2% (from 533.4 kWh to 571.7 kWh) under CS3. Significant economic savings were also demonstrated by the 21% reduction in SN energy costs in CS2. These results highlight how the AI-EMS may support the global shift to zero-carbon cities by coordinating HMGs within an SN, acting as a cloud-based ESS for the main grid, promoting energy system sustainability, resilience, and flexibility. • AI-EMS based on DDPG optimizes energy use across HMGs within the SN. • SN acts as cloud ESS, absorbing surplus energy and reducing imports from the main grid. • AI-EMS boosts grid stability by dynamically managing energy flows and storage. • AI-EMS was evaluated in three cases showing varying grid interactions and cost impacts. • Coordinated multi-HMGs maximize local resources, boosting profits and resilience.