Power Autonomous Allocation Strategy for Multi-Port Parallel Dual-Active-Bridge Converters Based on a Dynamic Consensus Algorithm

Conventional equal-sharing and fixed-capacity-weight allocation laws for parallel dual-active-bridge (DAB) converters do not explicitly account for time-varying module capacities, which can overload derated modules and under-utilize healthy ones. This paper presents a hierarchical derating-aware strategy for capacity-proportional power allocation. A shared DC-bus voltage controller supplies a common correction. Two dynamic average consensus (DAC) channels estimate the network-average power request and allocation capacity through neighbor communication; their ratio generates local capacity-proportional references, which are tracked by phase-shift controllers under common actuator constraints. A three-module averaged model is tested under load steps, module derating, communication delay, inductance/response mismatch, and input-voltage variation, with equal sharing, fixed capacity weights, and an ideal centralized capacity-aware reference as benchmarks. Local stability is analyzed for all nonzero graph modes and compared with continuous-delay and Padé models. For a three-module system sampled at 1 ms, the nominal maximum stable integer delay is 124 ms at a consensus gain of 4.2 s−1. In the tested abrupt-derating case, the proposed method restores the bus voltage to its ±2% band in 0.330 s, compared with 1.544 s for equal sharing and 1.490 s for fixed capacity weights. It meets the specified voltage and capacity-proportional sharing criteria in all 144 deterministic cases with a 10 or 80 ms delay. Actual output briefly exceeds the newly reduced capacity because estimation and power responses take time. These results demonstrate the simulation-level feasibility of capacity-proportional allocation for the tested connected undirected network with assumed capacity estimates.