Emergency mobile energy storage optimal allocation in microgrid-integrated distribution networks considering economic and resilience benefits

The accelerating pace of climate change has amplified the frequency and severity of extreme weather events, exposing power distribution systems to unprecedented vulnerabilities. Existing methods for emergency mobile energy storage (EMES) allocation often struggle to balance resilience enhancement and economic feasibility under large-scale disasters effectively. To address these challenges, this paper presents an advanced optimization framework for EMES deployment based on multi-agent Deep Reinforcement Learning (DRL). Specifically, a resilience curve model is developed to quantify system performance, focusing on resilience enhancement during post-extreme events. Subsequently, a novel Constrained Markov Nash Equilibrium Game (CMNG) model is established to address the constrained multi-objective optimization framework that integrates resilience enhancement and economic viability. To solve this model, a Multi-Agent Actor-Critic-Director (MAACD) algorithm employs a centralized attention aggregator architecture combined with the Lagrange multiplier method, transforming the constrained optimization problem into an unconstrained minimax problem to facilitate optimal decision-making. Comparative analyses demonstrate that the proposed framework achieves near-optimal solutions with resilience enhancement of 22.10 %, approaching the performance of the exact method I-AUGMENCON (22.97 %), while reducing computational time from 56.655s to 0.012s. Moreover, compared to other heuristic methods, MAACD exhibits superior performance in both solution quality and computational efficiency, achieving improvements of 1.73 %–7.97 % in resilience enhancement and 11.26 %–24.61 % in economic benefits while maintaining the lowest computational overhead. These results validate the effectiveness of the proposed methodology in achieving an optimized balance between solution quality and computational efficiency for real-time EMES deployment under extreme conditions.