Resumen de Detección y control del estado de una pila PEM para funcionamiento óptimo: arquitectura de agentes de percepción y control
Wilton Agila Gálvez
This manuscript presents a model for Fuel Cell Proton Exchange Membrane (PEMFC) real time state detection, based on a disturbance method and on a fuzzy decision tree classification method. The perturbations are applied, in different membrane humidification conditions, and the fuel cell electrical response is analyzed, to extract the best descriptors for the classification stage. The selected descriptors are the inputs of the classification model that is based upon a fuzzy decision tree, which is encapsulated as a state perception agent in a perception and control architecture based on expert agents. This state perception agent is integrated among other expert agents in fails detection and alert management, and in the efficiency control, in the real time perception and control architecture to get an autonomous operation of the fuel cell in the optimum hydration conditions. The architecture integrates perception and control algorithms that rely on sensors and context information and it is structured in a hierarchy of levels, each with a different temporal window and abstraction degree. These architectures are particularly suitable for complex non lineal systems, with strong interactions among parts, as it happens for the PEM fuel cell. The model displays the flexibility, versatility and ease of configuration required to face variations either in tasks, systems or scenarios (power requirement of a specific application). The work is multidisciplinary as it manages paradigms, techniques and models from many different areas, such as electrochemistry, electricity, electronics, computer sciences and artificial intelligence. The intelligent control system here proposed allows for the characterization and control of both cells with low power and medium-large stacks with high average power. The developed monitoring and control system for autonomous fuel cells operation has been demonstrated with different PEM fuel cells and functioning conditions displaying a high reliability in achieving the proposed energy efficiency target.
