Modeling and control of a fuel cell-battery hybrid vehicle

Richard T Meyer, Purdue University

Abstract

This work solves a fuel cell-battery hybrid vehicle (FCHV) power management problem using the embedding approach to hybrid (switched) optimal control problems. The FCHV is a switched system with three distinct modal configurations (modes): electric motor propelling/battery discharging, propelling/charging, and generating/charging. Each powertrain mode has a distinct set of dynamics and constraints. A supervisory-level switched system power flow model is developed as an interconnection of powertrain subsystems using component dynamical/algebraic models appropriate to power flow management. Given the model, a hybrid model predictive control strategy is set forth based on a minimization of a performance index (PI) that trades off velocity tracking error, battery state of charge variance, and electric drive and hydrogen fuel usages while penalizing frictional braking to encourage regenerative braking. The embedding approach representation of the optimization problem is solved, after applying collocation, with MATLAB's fmincon to compute mode switches and continuous time controls. To demonstrate the effectiveness of the methodology, an example mid-size sedan FCHV is simulated using five driving velocity profiles, which include three regulatory profiles. Drive cycle fuel consumption is shown to be lower than that from the Equivalent Consumption Minimization Strategy, a popular power management strategy. Finally, mode and continuous control projection are required in the infrequent case when the embedding approach solution does not result in a switched system solution; the projection methods used during the simulations are justified with an empirical study. The study's goal was to identify which of the eight mode and control projection combinations tested results in the least error between the embedding approach solution and simulated switched system costs. This work also demonstrates the superiority of the embedding approach over alternative methods for solving hybrid optimal control problems. The embedding approach is compared to multi-parametric programming, mixed-integer programming, gradient-descent based methods, and CPLEX in the context of five published examples: a spring-mass system, moving-target tracking for a mobile robot, two-tank filling, DC-DC boost converter, and a skid-steered vehicle. Results from the examples make clear the advantages of the embedding approach: lower performance index cost, generally faster solution time, and convergence to a solution when other methods fail.

Degree

Ph.D.

Advisors

DeCarlo, Purdue University.

Subject Area

Electrical engineering|Mechanical engineering

COinS