GOALI: Fast Nonlinear Model Predictive Control for Dynamic Real-time Optimization

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

1160014-Biegler For over three decades, Real-Time Optimization (RTO) and Model Predictive Control (MPC) have emerged as essential technologies for optimal process operation in the chemical and refining industry. More recently, MPC has been extended to Nonlinear Model Predictive Control (NMPC) in order to realize high-performance control of highly nonlinear processes. Moreover, for many applications there is a need for RTO to evolve from steady-state optimization models to dynamic models, especially for systems, such as batch and cyclic processes, that are never in steady state. Both NMPC and dynamic real-time optimization (D-RTO) allow the incorporation of first principle process models, which lead to on-line optimization strategies consistent with higher-level tasks, including scheduling and planning. However, more detailed dynamic optimization models that reflect complex reaction and separation phenomena and multi-stage dynamic operation still need to be addressed ? and solved as time-critical, on-line applications. Here, a major concern is that computational times needed to solve these large-scale optimizations lead to feedback delays in implementation that can degrade performance and possibly destabilize the process. This proposal addresses these issues and furthers the realization of fast on-line dynamic optimization with first principle models. Our previous work led to a class of sensitivity-based algorithms that separate dynamic optimization into background calculations, where most of the computation is performed, and online calculations, where a perturbed problem is solved very quickly. On-line computations are thus reduced by several orders of magnitude and become very fast, even for large, complex nonlinear models. These formulations were developed both for NMPC as well as state and parameter moving horizon estimation (MHE). The intellectual merit of the proposed activity extends the development and analysis of sensitivity-based on-line optimization with first principle dynamic models, particularly advanced-step NMPC and MHE. This transformative proposed work leads to nonlinear model predictive control and on-line dynamic optimization for large-scale chemical processes without the limitations of computational feedback delay. Advances will be developed in the solution of background NLPs over multiple sampling times. In addition, we propose to extend advanced-step NMPC and MHE to hybrid systems, where discrete decisions (switches) are allowed at any point in time, and the algorithm suffers no loss in computational efficiency. Moreover, we will incorporate reduced order nonlinear dynamic models, develop specialized NMPC and MHE approaches for these problems and extend them to dynamic real-time optimization. Broader Impacts: Broader impacts resulting from the proposed activity include the development and application of this approach to a number of challenging power generation processes. Characterized by load changes and dynamics with strong nonlinearities, performance of these multi-stage systems can be greatly improved through efficient NMPC and MHE strategies. These concepts will also be integrated within a comprehensive real-time optimization framework that combines open source optimization and sensitivity codes with a state of the art modeling environment. Finally, graduate training is emphasized as a key component of this proposal. Included in the educational plan are industrial interactions with GE Global Research and the development of courses and materials related to Dynamic Real-Time Optimization.