Mechanism-data hybrid driven dual-mode ADMM-EMPC for dynamic economic optimization of boiler-turbine units

To enhance dynamic economic performance of flexible-load boiler-turbine units while preserving stability, this paper proposes a mechanism-data hybrid driven dual-mode ADMM-EMPC. The hybrid model embeds data-driven nonlinear characterizations of key mechanism-related quantities into mass- and energy-balance equations, retaining physical interpretability and control-oriented differentiability. An ADMM-based auxiliary controller further exploits economic potential during stability recovery under Lyapunov constraints, while a candidate-gain acceptance and fixed-gain fallback mechanism improves reliability under finite real-time computation. Validation using plant DCS data and simulation-based closed-loop tests demonstrates accurate modeling, robust behavior under off-nominal constrained conditions, and improved control performance. Under nominal continuous AGC tracking, the proposed ADMM-EMPC reduces the economic performance index by 10.4% relative to conventional MPC, while further reducing unit-load RMSE by 41.6% relative to fixed-gain EMPC.