Genetic Programming Control System Design for Low Temperature PEM Fuel Cell

  • Abdel Gafoor Haddad

Student thesis: Master's Thesis

Abstract

A quick and accurate control system for the air supply in fuel cell systems (FCS) is required to prevent oxygen starvation or to maximize the net power. For this purpose, regulating and gain-scheduled controllers are designed by employing genetic programming (GP) and an artificial neural network (ANN). To speed up the convergence of GP, order reduction of the FCS models is carried, and a comparison between the models is performed. A fifth-order model is selected for the purpose of control design due to its quick run time with a minimal loss in response accuracy of the oxygen excess ratio (OER). Guidelines on applying GP based on data obtained from simulations are developed. The overfitting phenomenon is observed, and several methods to overcome it are proposed. The linearizing property of GP is enforced through the specially designed cost function to ensure consistent performance over a wide range of operating conditions. The cost function is selected based on the minimization of OER integral absolute error, excursion, or both. The generated GP controller is compared to the dynamic feed-forward with PI controller. Adaptation is added to the OER regulation problem by training an ANN that provides the optimal OER reference according to the stack current and temperature. The performance of both the regulation and gain-scheduled controllers is tested under noise in the compressor flow and the stack current measurements. The robustness of the GP controllers is analyzed through linear frequency response analysis. The describing function of multiple controllers is obtained numerically and is used to generate the Bode plots of the closed loop systems at different input amplitudes. To show the linearizing capability of GP, variations in the Bode plots of GP-based control systems is compared to traditional control systems.
Date of AwardMay 2020
Original languageAmerican English

Keywords

  • Genetic programming
  • fuel cell
  • oxygen starvation
  • system linearization
  • optimal oxygen excess ratio.

Cite this

'