Hierarchical Distributed Control With Projected Gradient Descent for Real-Time Congestion Mitigation in MVDC Distribution Systems

Ineffective congestion management in medium-voltage DC distribution system risks thermal overloads, voltage violations, and systemic instability. To address this challenge, this paper proposes a distributed congestion mitigation method via adaptive real-time control. First, a hierarchical control framework integrating congestion constraints is established, with optimality conditions rigorously derived. Second, the upper layer generates dispatch references to guide lower-layer operations. Projection operators are then introduced to reformulate optimality conditions for controller compatibility, enabling distributed optimal control laws based on projected gradient descent. This approach achieves real-time congestion management and optimal operation. Simulations and comparisons are conducted on a modified IEEE 33 medium-voltage DC distribution system. Compared to model predictive control, the proposed method achieves reduced fluctuations under varying operating conditions. In contrast to distributed robust optimization, it exhibits superior economic performance.