Investigation of the use of transient process data for steady-state Real-Time Optimization in presence of complex dynamics

Curvelo, Rodrigo; Delou, Pedro de Azevedo; Souza, Maurício B. de; Secchi, Argimiro Resende

doi:10.1016/b978-0-323-88506-5.50200-x

Cited by 4 publications

(2 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Overall, our study demonstrates the potential benefits of residual REINFORCE to tune the inputs calculated by the SSRTO online using transient measurements in the absence of a dynamic mechanistic model. This is similar to the methodologies developed by Delou et al, Curvelo et al and Demuner et al, which employed a strategy centered on a Hammerstein model structure in which an identified process steady-state model represents its static nonlinear function. Our approach presents similarities to the methodologies developed in Hassanpour et al, , which created a reinforcement learning-based process controller design that can be practically implemented.…”

Section: Resultsmentioning

confidence: 90%

Gas-Lift Optimization Using Physics-Informed Deep Reinforcement Learning

de Rezende Faria,

Capron,

Secchi

et al. 2024

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

Real-time optimization (RTO) methodologies have become essential for optimal process operation in the oil and gas industries. Typically, RTO is based on a steady-state model (steady-state real-time optimization�SSRTO) and operates as a closed-loop optimizer. However, this technique can result in suboptimal policies due to steady-state waiting and plant-model mismatch issues. To alleviate these problems, we combine this closed-loop optimizer with a data-driven residual optimizer based on deep reinforcement learning (DRL). The idea is to use the SSRTO solution's stability and convergence properties while reducing plant-model mismatch through a residual DRL controller that can tune the optimal inputs calculated by the SSRTO based on transient measurements (i.e., real data). Our proposed methodology delivered robust online control policies compared to dynamic real-time optimization (DRTO) for closed-loop control, even when the plant-model mismatch was simulated by introducing parameter uncertainty and noise in the plant and using an SSRTO model with structural uncertainty. As a result, its economic gain was improved compared to the SSRTO solution, with an average computational cost 11 times lower than DRTO, making it suitable for online applications.

show abstract

Section: Resultsmentioning

confidence: 90%

Gas-Lift Optimization Using Physics-Informed Deep Reinforcement Learning

de Rezende Faria,

Capron,

Secchi

et al. 2024

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

show abstract

“…Shamaki and Odloak (2020) combinaram a ROPA com um controlador preditivo de horizonte infinito ( infinite horizon zone control MPC). Curvelo et al (2021) estudaram a performance da ROPA quando aplicada a diferentes padrões de estados transientes, como mudanças rápidas ou lentas no comportamento do sistema, e verificaram a melhor performance da ROPA na maioria dos padrões. Por último, Matias et al (2020) propuseram estender o contexto da ROPA para otimização de uma planta utilizando a estimação assíncrona de parâmetros, essa abordagem foi explorada na introdução.…”

Section: Objetivosunclassified

Otimização em tempo real aplicada em um sistema experimental de Gas Lift.

Oliveira¹

View full text Add to dashboard Cite

Real-time Optimization (RTO) is a production optimization technique that aims at improving plant economic performance in real-time. Many proposals of this technique have been made, being the two step RTO known as model parameter adaptation (MPA) the most traditional. Although being widely used in industrial application, the MPA uses stationary process information to perform the optimization and is not appropriated for dynamic systems. Recently a real time optimization with persistent adaptation (ROPA) has been proposed as a new RTO approach to handle and optimize dynamic systems. With a hybrid approach that uses stationary and transient information this method has shown promising results, but more studies are still necessary to support those results. The absence of a real system application, only implementation in simulated environments, and the lack of studies in different approaches regarding this method are some of the issues that still needs to be discussed. To contribute to the development of ROPA, this work presents an implementation of ROPA on an experimental rig that emulates a subsea oil extraction by Gas Lift. This experiment will evaluate ROPA performance in an real environment and compare the results with MPA. In order to apply ROPA two different online estimators are selected, the extended Kalman filter (EKF) and the moving horizon estimation (MHE). This work begins by modelling the experiment, since MPA and ROPA are model based, then MPA and ROPA-EKF are implemented and compared. Finally, ROPA-MHE is implemented and compared with ROPA-EKF. In order to implement ROPA-MHE a study on the arrival cost, a penalty term of MHE, was performed and it was verified the importance of this term on obtaining accurate estimations. The results obtained in this work have shown that ROPA-EKF had a better performance than MPA, specially in transient states, and ROPA-EKF and ROPA-MHE had similar performance.

show abstract