(56d) Classification of Empirical Models with Profit Considerations

Most real-world industrial processes are highly complex and are increasingly modeled using empirical system identification techniques. However, these empirical models are difficult to interpret from a scientific standpoint, consequently making it difficult to predict system behavior or develop effective control strategies for the process ^[1]. Hence, it is desirable to develop models that are easily understood by humans. This talk draws inspiration from ideas first introduced in ^[2,3,4] where a process was operated at conditions that prioritized the collection of data to characterize the dynamics of a process and use economic model predictive control (EMPC) ^[5] to ensure process stability and feasibility. The flexibility of economic model predictive control (EMPC) was taken advantage of, along with Lyapunov-based EMPC (LEMPC) ^[6] constraints to develop explicit guarantees of stability.

This talk will focus on the development of an automated data-gathering technique to classify empirical models as being ``physics-based'' or not based on human understanding of first principles while making guaranteed profit considerations. The contribution that this talk will focus on is the development of an automated implementation strategy that uses learning algorithms to identify new states to force an empirically modelled process to collect information vital to characterizing it as being physics-based or not. The formulation takes advantage of multiple redundant optimization-based model predictive control strategies to determine if a process operating in a safe region of operation can be manipulated to a non-optimal state to collect data required for identification without losing significant profits. Two LEMPC formulations with slight changes in their constraints will be used to guide a third control formulation to force the real process to in this implementation strategy. Economic guarantees are incorporated into the implementation of the control strategy to quantify the impact of data-collection required to better understand processes and make explicit guarantees of stability and profitability. A process example will be discussed to exemplify the strategy.

References

[1] Alanqar, Anas, Matthew Ellis, and Panagiotis D. Christofides. "Economic model predictive control of nonlinear process systems using multiple empirical models." In 2015 American Control Conference (ACC), pp. 4953-4958. IEEE, 2015.
[2] L. Giuliani and H. Durand, “Data-based nonlinear model identification in economic model predictive control,” Smart and Sustainable Manufacturing Systems, vol. 2, pp. 61–109, 2018.
[3] H. Oyama, D. Messina, K. K. Rangan, and H. Durand, “Lyapunov-based economic model predictive control for detecting and handling actuator and simultaneous sensor/actuator cyberattacks on process control systems,” Frontiers in Chemical Engineering, vol. 4, p. 810129, 2022.
[4] H. Oyama, A. F. Leonard, M. Rahman, G. Gjonaj, M. Williamson, and H. Durand, “On-line process physics tests via lyapunov-based economic model predictive control and simulation-based testing of image-based
process control,” in 2022 American Control Conference (ACC). IEEE, 2022, pp. 2479–2484.
[5] M. Ellis, H. Durand, and P. D. Christofides, “A tutorial review of economic model predictive control methods,” Journal of Process Control, vol. 24, pp. 1156–1178, 2014.
[6] M. Heidarinejad, J. Liu, and P. D. Christofides, “Economic model predictive control of nonlinear process systems using lyapunov techniques,” AIChE Journal, vol. 58, no. 3, pp. 855–870, 2012.