Nonlinear model predictive control of a multistage evaporator system using recurrent neural networks

摘要: The use of multistage evaporators, motivated by the energy economy from reusing flashed steam is common in a wide range process industries. Such evaporators however present several control problems which manifest form strong interactions among many variables, significant dead times, tendency to open-loop instability and severe nonlinearities. In this paper, nonlinear model predictive (NMPC) scheme utilizing proportional-integral (PI) controller its inner loop developed for simulated industrial-scale five-stage evaporator using continuous-time recurrent neural network state space as internal model. Input-output data obtained closed-loop system identification experiments are used training Levenberg-Marquardt algorithm with automatic differentiation. A similar approach developing an optimal law plant based on predictions. effectiveness tested simulating various problem scenarios involving set-point tracking disturbance rejection comparing performance that decentralized PI controllers earlier. Results show improvements performance, particularly terms settling time.