Short summary of the project

This project aims to provide innovative technologies to bridge the gap between Model Predictive Control (MPC) and artificial intelligence (AI). It outlines the challenges of traditional MPC applications, emphasizing the need for digitalization and AI. The research objectives center on bridging the gap between MPC and AI through developing Hybrid-AI-MPCs, leveraging intelligent control algorithms and advanced modeling to improve efficiency, productivity, and control performance in the energy industry, ultimately contributing to a sustainable future.

Project objectives

Provide robust methodologies for building and applying robust AI-based process controllers, with stability assurance and uncertainty assessment through Bayesian inference mechanisms.

Scope

• This project aims to develop a hybrid control strategy between the fundamentals proposed by the nonlinear predictive controller and the artificial intelligence models. The controller will use artificial intelligence techniques to predict the model and obtain the closed-loop solution. In other words, the controller will be built as a closed-loop model of a predictive controller that, given the process conditions and the desired operating points, will provide the trend of actions and control that must be inserted in the plant to stabilize, minimize the disturbances and bring the system to the desired operating point.
• The main known limitations for the construction of the proposed controller are:
• The computational cost associated with closed-loop system training can be prohibitive. However, new training and parallelization techniques have been proposed, and despite being outside the scope of the current thesis, they are advances that may contribute to the completion of the work.
• Real-time implementation in process systems. In this regard, the main challenge is to embed this technology in systems with insufficient computational capacity to make predictions in operating time.

Innovation potential:

• Hybrid control strategies.
• Robust AI-based process controllers.