Adaptive congestion management for EV charging in synchrophasor enabled ADNs: A decentralised distributed approach with concordance framework

The widespread expansion of Electric Vehicles (EVs) and the integration of charging stations into Active Distribution Networks (ADNs) present unprecedented challenges and opportunities for grid management. The paper presents a novel coordinated congestion management framework, which effectively integrates Long Short-Term Memory (LSTM) networks with a modified Deep Deterministic Policy Gradient (DDPG) algorithm enhanced by the Gurobi solver, termed LGDDPG to achieve adaptive, event-aware EV charging. A modified IEEE 13 bus radial distribution system was examined, with Electric Vehicle Supply Equipment (EVSEs) connected to the buses at Grid Connection Points (GCPs). Each EVSE acts as a decentralised reinforcement learning (RL) agent and functions as a cluster of electric vehicle loads that autonomously adjusts its charging rate by utilising PMU-based index data with inter-agent communication. The proposed test environment is exposed to diverse event scenarios, enabling the RL agent to effectively explore the state space, thereby testing the adaptability of the system. To evaluate the economic viability of the proposed framework, a dynamic congestion tariff was introduced, and its performance was systematically benchmarked against conventional pricing structures and alternative RL approaches, thereby demonstrating both its effectiveness and robustness. This concordance-distributed intelligence and fairness mechanism guarantees the equal allocation of available grid capacity among EVSEs, as measured by Jain index at the aggregated system levels. Significant findings from exploratory data analysis demonstrate a 97.5% success rate and narrow variability in cost analysis for the proposed framework when evaluated with other baseline methods, demonstrating substantial improvements in the effectiveness, fairness, and economic advantage of adaptive charging. • Adaptive ADN congestion management is done via dynamic adjustment of EVSE power. • LGDDPG framework enhances fair allocation of power and satisfies constraint limits. • SGS Index enables grid observability and awareness via synchrophasor feedback. • Dynamic congestion tariff is effective in comparison to other pricing models.