Digital twin enhanced hybrid modelling for main steam temperature optimization of flexible power plant

Under the dual operational stresses of thermal storage system integration and rapid peak-shaving demands, the main steam system confronts escalating disturbances, amplified dynamic deviations, and prolonged regulation cycles. To address these challenges, this study proposes a digital twin-based hybrid modeling approach for optimized main steam temperature control, synergistically combining mechanism modeling, deep learning, and intelligent optimization algorithms to significantly enhance the dynamic performance of control systems. The methodology comprises three key innovations: First, a high-precision dynamic mechanism model is established through multi-physics coupling analysis of flow and heat transfer processes in the main steam system. Second, a dynamic heat absorption soft-sensing model and multi-parameter collaborative optimization framework are developed, overcoming the limitations of conventional static heat absorption characterization to achieve 54.4 % improvement in model accuracy. Third, an advanced control strategy incorporating multi-step predictive optimization of dynamic heat absorption is implemented. Simulation results demonstrate remarkable improvements compared with conventional PID control: PeakTime and TransientTime are reduced by 58.49 % and 68.15 % respectively. Under load variation scenarios, steam temperature fluctuations are dramatically suppressed from 26.9 °C to merely 0.2 °C. This research provides an innovative solution for flexible peak regulation and operational safety of coal-fired power units, demonstrating significant engineering application value.