This paper proposes an offset-free Quasi-Infinite Horizon Nonlinear Model Predictive Controller (QIH-NMPC) using online parameter adaptation. In the proposed method, the adaptation law is modeled by a first order differential equation as a function of the tracking error and subsequently combined with a QIH-NMPC algorithm for online updating of the unknown parameter. The effectiveness of the proposed control scheme is demonstrated on a continuous stirred tank reactor (CSTR) and an experimental cascaded three-tank system with uncertain model parameters, structural plant/model mismatch and noisy measurements. For the purpose of comparison, the state-of-the-art online state and parameter estimators such as moving horizon estimation (MHE) and extended Kalman filter (EKF) were also incorporated into QIH-NMPC algorithm. The simulation and experimental results obtained showed the efficacy of the proposed adaptation scheme as it demonstrated a comparable performance to standard estimators (MHE and EKF) although with a lesser computational time.
Stable and offset-free responses are two desirable properties of nonlinear model predictive control (NMPC). However, traditional long short-term memory (LSTM)-based NMPC lacks inherent stability guarantees, especially in the presence of process-model mismatch. To address this limitation, this study integrates a Lyapunov stability constraint and an integrated output disturbance model into LSTM-based NMPC to enforce closed-loop stability and offset-free performance. Two simulation examples are used to demonstrate the effectiveness of the proposed method: the continuous stirred tank reactor (CSTR) and the forced evaporator process. LSTM models were trained using input-output datasets collected from the two processes and were subsequently employed as prediction models within the NMPC framework. The designed LSTM-based NMPC was then implemented on simulation models of the processes. Validation results for the CSTR process showed that the trained LSTM model achieved excellent fits of 93.72% and 95.12% for concentrations C A and temperature T , respectively, while the linear model yielded poor fits of 9.37% and 19.88%. For the forced evaporator process, the LSTM model also achieved high prediction accuracies with fits of 96.76%, 96.52%, and 85.69% for L 2 , X 2 , and P 2 , respectively, whereas the linear model produced significantly lower fits of 51.31%, 27.69%, and 2.24%. Closed-loop simulation results demonstrated that the proposed LSTM-based NMPC exhibited stable and offset-free setpoint tracking in spite of process/model mismatches, performing comparably to NMPC based on first-principles (FP) models. However, it significantly outperformed the linear model predictive control (LMPC).