The present project aims, on the one hand, to develop advanced tools to apply the multiobjective optimization methodology in the design of control systems for complex multivariable processes and, on the other hand, to advance in the development of multivariable controllers using multiobjective optimization, with special emphasis on model-based predictive controllers (widely used in the process industry). Additionally, all results will be validated in relevant applications in various engineering fields, mainly energy and transportation. In particular, work will be carried out on systems based on PEM fuel cells (both for static and mobile applications) and on aspects related to the search for optimal operating points, energy management in vehicles, trajectory control, etc.
In its scientific facet, the project will seek to provide novel solutions in the three fundamental components of a multi-objective optimization problem: in the objective selection phase, bringing them closer to the designer’s needs; in the optimization phase, improving the current algorithms to deal with a large number of variables and constraints; and in the decision support phase, improving the systems for incorporating the designer’s preferences and multidimensional visualization.
In addition, applying the developed multiobjective tools to the design of multivariable controllers, the aim is to achieve an intuitive and interpretable way of adjusting these controllers, especially in the case of control problems with many constraints. Specifically, model-based predictive control or MPC will be used as a control methodology, which, besides being a technology with great industrial acceptance, is closely related to optimization. In parallel, given the importance of virtual models and sensors when complex multivariable processes are to be controlled, the multi-objective methodology will also be applied to achieve advances in the quality and reliability of dynamic models and in the development of virtual sensors.
SPECIFIC OBJECTIVES
Scientific objectives:
Development of new multi-objective optimization algorithms and/or variants of existing ones to improve performance in problems with many variables and many constraints.
Development of techniques and procedures that, incorporated to the optimization algorithms, increase the relevance in the set of solutions obtained.
Development of tools for comparison of multidimensional Pareto fronts and their incorporation in an interactive decision support software oriented to control engineering.
Design of aggregators or new preference incorporation methodologies for the decision support system.
Development of new indicators/targets to enable the use of multi-objective methodology in MPC model identification, controller parameter tuning and reference management.
Design of reliable virtual sensors in the face of uncertainties or errors in the models and uncertainties in the treatment of constraints using heuristic optimization techniques.
Development of simplified p-step prediction models for incorporation into online MPC controllers.
Design of multi-objective predictive controllers, with application of multi-objective optimization methodology and automation of the decision making phase.
Technological objectives:
Develop a complete nonlinear dynamic model of the propulsion system of a vehicle based on PEM stack (fuel cell, batteries, DC bus, brushless motor).
Development of a Hardware-In-the-Loop (HIL) system for preliminary testing and validation of controllers for the two platforms mentioned above.
Development and implementation of advanced optimization and control algorithms for energy management of fuel cell based systems (Energy Management Module) with the corresponding adaptations to each of the two test platforms.