Advanced delay-aware two-stage independent control for SOFC CCHP systems against complex gas transport

Existing 0D dynamic models of solid oxide fuel cell (SOFC) systems typically neglect gas transport delay. In this study, an SOFC combined cooling, heating, and power (CCHP) system integrating a double-effect absorption chiller is developed. Soft and hard delay parameters are introduced to characterize the gas transport in auxiliary components. A two-stage independent control framework, combining Kalman filter-adaptive model predictive control (KF-AMPC) and integral compensator-deep reinforcement learning (IC-DRL), is implemented for temperature regulation and power tracking. Comparative and sensitivity analyses show that soft delays produce cumulative lag along the gas-flow direction, whereas hard delays introduce asynchronous responses at different locations. In the upstream subsystem, AMPC reduces the hotbox-temperature MAE by 93% compared with fixed-model MPC over the investigated operating range, while KF state estimation suppresses noise-induced high-frequency control oscillations. In the downstream subsystem, IC-DRL achieves the lowest MAE, RMSE, and MaxAE among the compared controllers. Large or rapid load variations can still induce temporary instability. A reference update interval of 5 min or less is recommended for upstream power-temperature control, whereas 10 min or longer is preferred for downstream exhaust gas-cooling control.