Real-Time Energy Management of PV-ESS Integrated Active Distribution Networks Using Digital Twin-Enabled Deep Reinforcement Learning
The increasing penetration of photovoltaic (PV) generation and battery energy storage systems (BESSs) has significantly increased the operational complexity of active distribution networks, where real-time energy management must simultaneously address renewable uncertainty, voltage regulation, and battery lifetime preservation. Existing Digital Twin-based energy management approaches primarily support monitoring and visualization, whereas deep reinforcement learning (DRL) controllers are commonly developed independently of real-time system synchronization, limiting their adaptability under rapidly changing operating conditions. To overcome these limitations, this paper proposes a Digital Twin-enabled Deep Reinforcement Learning (DT-DRL) framework for coordinated PV–BESS energy management in active distribution networks. The proposed framework establishes a closed-loop cyber–physical architecture in which continuously synchronized Digital Twin states are directly incorporated into a Proximal Policy Optimization (PPO)-based decision-making process. A multi-objective formulation is developed to jointly minimize operating cost, voltage deviation, and battery degradation while satisfying network operational constraints. Renewable generation and load uncertainties are represented using Monte Carlo-based stochastic scenarios to improve policy robustness under practical operating conditions. The proposed framework is validated on the IEEE 33-bus distribution system and compared with rule-based control (RBC), optimal power flow (OPF), and conventional DRL approaches. Simulation results demonstrate that the proposed method reduces the daily operating cost by 22.2%, decreases the maximum voltage deviation to 0.039 p.u., and achieves more stable BESS operation with lower operational variability under uncertain conditions. Furthermore, the complete Digital Twin synchronization and PPO decision-making process requires only 0.41 s per control interval, satisfying the timing requirements of distribution-level energy management systems. These results demonstrate that the proposed DT-DRL framework provides an accurate, computationally efficient, and practically deployable solution for real-time energy management in renewable-rich active distribution networks.