10.1184/R1/6720857.v1 Rui Huang Rui Huang Nonlinear Model Predictive Control and Dynamic Real Time Optimization for Large-scale Processes Carnegie Mellon University 2010 Chemical Engineering 2010-12-01 00:00:00 Thesis https://kilthub.cmu.edu/articles/thesis/Nonlinear_Model_Predictive_Control_and_Dynamic_Real_Time_Optimization_for_Large-scale_Processes/6720857 <p>This dissertation addresses some of the theoretical and practical issues in optimized operations<br>in the process industry. The current state-of-art is to decompose the optimization<br>into the so-called two-layered structure, including real time optimization (RTO) and advanced<br>control. Due to model discrepancy and inconsistent time scales in different layers,<br>this structure may render suboptimal solutions. Therefore, the dynamic real time optimization<br>(D-RTO) or economically-oriented nonlinear model predictive control (NMPC)<br>that directly optimizes the economic performance based on first-principle dynamic models<br>of processes has become an emerging technology. However, the integration of the firstprinciple<br>dynamic models is likely to introduce large scale optimization problems, which<br>need to be solved online. The associated computational delay may be cumbersome for the<br>online applications.<br><br>We first derive a first-principle dynamic model for an industrial air separation unit (ASU).<br>The recently developed advanced step method is used to solve both set-point tracking and<br>economically-oriented NMPC online. It shows that set-point tracking NMPC based on the<br>first-principle model has superior performance against that with linear data-driven model.<br>In addition, the economically-oriented NMPC generates around 6% cost reduction compared<br>to set-point tracking NMPC. Moreover the advanced step method reduces the online<br>computational delay by two orders of magnitude.<br><br>Then we deal with a realistic set-point tracking control scenario that requires achieving<br>offset-free behavior in the presence of plant-model mismatch. Moreover, a state estimator<br>is used to reconstruct the plant states from outputs. We propose two formulations using<br>NMPC and moving horizon estimation (MHE) and we show both approaches are offsetfree<br>at steady state. Moreover, the analysis can be extended to NMPC coupled with other<br>nonlinear observers. This strategy is implemented on the ASU process.<br><br>After that, we study the robust stability of output-feedback NMPC in the presence of plantmodel<br>mismatch. The Extended Kalman Filter (EKF), which is a widely-used technology<br>in industry is chosen as the state estimator. First we analyze the stability of the estimation<br>error and a separation-principle-like result indicates that the stability result is the same as<br>the closed-loop case. We further study the impact of this estimation error on the robust<br>stability of the NMPC.<br><br>Finally, nominal stability is analyzed for the D-RTO, i.e. economically-oriented NMPC,<br>for cyclic processes. Moreover, two economically-oriented NMPC formulations with guaranteed<br>nominal stability are proposed. They ensure the system converges to the optimal<br>cyclic steady state.</p>