Loading...

Optimal Design and Intelligent Control of Polymer Electrolyte Membrane Fuel Cell Stack

Ahmadi, Mohammad Reza | 2022

71 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 55869 (46)
  4. University: Sharif University of Technology
  5. Department: Energy Engineering
  6. Advisor(s): Boroushaki, Mehrdad
  7. Abstract:
  8. We present here an analysis of controlling the Polymer Electrolyte Membrane Fuel Cells (PEMFCs) using the Q-learning algorithm, the most widely-known among reinforcement learning (RL) techniques. The method is to train the controller to guide and sustain the fuel cell power output in the 2.5 kW mark by way of manipulating elements of the reaction subsystem including the fuel cell current, the relative humidity, and the anode/cathode pressures. As the Q-learning algorithm need be implemented within a fuel cell simulation environment, the mathematical model known as Amphlett steady-state model of the PEM fuel cell was employed. The semi-empirical nature of this model necessitates the extraction of numerical values for certain parameters of the model based on a system serving as the Reference. This was achieved through implementation of the genetic algorithm (GA) and the particle swarm optimization (PSO) technics, the goal being to obtain best values—in the sense of maximal compatibility—with regards to the Reference. Upon learning, the now-reinforced controller was set through a series of 15 consecutive test runs for performance assessment. The results, comprised of diagrams of instant reward and power output, indicate a highly desirable potential for sustaining fuel cell power. This project also aims at the improvement of maximum power output attainable. Two distinct variables were thus studied: a. the height to width ratio of the inlet to stack channel of the reaction gases (as a geometrical characteristic of the fuel cell); and, b. stoichiometric factors relevant to the reactant flows (as an electrochemical characteristic affecting max attainable power)—best values were obtained through simulation via COMSOL Multiphysics. A comparison shows 9.97% increase in the max power output over the Reference
  9. Keywords:
  10. Reinforcement Learning ; Proton Exchange Membrane (PEM)Fuel Cell ; Maximum Power Output ; Polymer Electrolyte Membrane (PEM) ; Q-Learning ; Power Output ; Power Output Control

 Digital Object List

 Bookmark