Integrating Dynamic Real-time Optimization And Model Predictive Control For Two-layered Large-scale Processes

Global optimization and control of large-scale industies are organized in a hierarchical structure to decompose the unmanageable problem into a cascade of interconnected solvable problems.Each layer consider its issues in different time scales.RTO is implemented where economically justified and is typically formulated based on a profit function of the plant,computing optimal set-point.APC works to track the set-point,taking constraints and disturbances into account.In practice,fluctuations in planning and scheduling,raw material prices,market needs and plant models would lead system operating ponits to change frequently,in which situation,economic performance in dynamic process really matters.Traditional steady-state RTO lacks consideration in dynamic economic performance.To improve system economic performance in the whole operating period,this paper designs dynamic optimization and control strategy,integrating D-RTO and MPC.On this foundation,event-triggered D-RTO based on economic performance is proposed and applied to a typical large-scale industry.Firstly,in the context of hierarchical structure,a dynamic real time optimization algorithm is proposed.The upper layer utilizes EMPC(Economic MPC)to compute economically optimal dynamic trajectory with a series of feasible steady-state operating points,these points are reachable for the lower layer.MPC is utilized in the lower layer to track the mentioned steady-state operating points.The feasiblity and stability of the proposed algorithm is analyzed.Simulation on CSTR verify its validity.Then,we propose a two-layered control framework integrating event-triggered dynamic real-time optimization and model predictive control,addressing the issue of infeasibility and performance loss caused by uncertainties.The upper layer utilizes EMPC to minimize an economic cost function,calculating reference trajectories.The lower layer utilizes Lyapunov-based model predictive controller to steer the plant to track the reference trajectories.A triggering criterion based on cost function is proposed to compensate for performance loss.When the deviation between these two layers' cost function exceeds a pre-set threshold,the upper layer is triggered to execute optimization and update the reference trajectory based on the current system states.Feasibility and closed-loop stability of the proposed two-layered control framework has also been analyzed.Simulation on CSTR verify its validity.Finally,considering model parameter uncertainties,the algorithm integrating eventtriggered dynamic real-time optimization and model predictive control is applied to reaction and regenerator system of fluid catalytic cracking units in oil refining process.Economic performance is improved by this algorithm.We set different weights and thresholds to analyze how triggering criteria influences system performance.The experiment demonstrates that this proposed algorithm can not only make system operate economically,but also make compromise between performance in different aspects according to control requirements.