Modeling and Optimization of Fuel Cell Electric Vehicle Considering Wide Variation of DC Link Voltage to Electric Powertrain and Air Supply System

Abstract

This paper dealt with detailed fuel cell electric vehicle (FCEV) modeling and fuel economy maximization based on surrogate modeling. The developed FCEV model considered the electric powertrain and air supply system. And optimization was conducted accounting for these components.

The contribution of this paper is below. 1) Air supply system was modeled and reflected to fuel cell stack performance and efficiency 2) Detailed FCEV model was developed based on real driving test results 3) Reflect wide voltage variation for traction motor and air compressor motor by using the artificial neural network (ANN) 4) Adaptive layering and sampling (ALS) algorithm was suggested

First of all, the FCEV modeling was performed. From the real driving condition, the driving condition was generated by assuming a grade profile from real driving data. During the modeling process, to secure the fidelity of the developed model, real driving test data was reflected as much as possible. For the traction motor modeling, the actual dimension of the cross-sectional area of the rotor and the stator, the number of turns and coil specification of armature winding, and material properties were utilized. Based on these, 2-dimensional (2D) electromagnetic finite element analysis (FEA) was conducted and predicted the performance and efficiency of the traction motor. For the air supply system, test result-oriented modeling was performed. To do this, using the air compressor prototype, an air compression test was conducted for various speeds and inlet valve angles. In addition, the air compressor controller was modeled based on air compressor motor speed real driving test data. At this time, utilize the optimization technique, undetermined controller gains were determined to minimize the normalized root mean squared error (NRMSE) between the simulation profile and the real driving test profile. The fuel cell model was adopted from the MATLAB Simulink library, and the battery was modeled as the RC equivalent model based on the test result. The fuel cell model was chosen as a detailed model, which can reflect supplied air flow and pressure.

Next, simulation and optimization of FCEV were conducted. For that, the design problem formulation was preceded. Then, ANN surrogate modeling was performed to conducting simulation and optimization considering voltage variation. To simulate the traction motor and air compressor motor considering design variable and voltage variation, ANNs were constructed for these two motors. The input of the ANN was the stack length, armature turns, DC voltage, speed, and torque for each motor and the output of the ANN was the maximum achievable torque and total loss. The generated ANN for two motors were planted to the FCEV model, and FCEV ANN was constructed. The input of the ANN was the stack length, armature turns of each motor, and gear ratio and the output of the ANN was fuel economy of FCEV and driving cycle trace NRMSE. Finally, based on this generated FCEV ANN, fuel economy maximization was achieved considering trace NRMSE. In this part, the adaptive layering and sampling algorithm (ALS) was proposed. It allows determining the number of hidden layers and samples for ANN construction.

Finally, to confirm the developed FCEV model, the simulation profile was investigated for each part. In addition, the effectiveness of ANN and the suggested ALS algorithm was validated from simulation and modeling results. And for the three driving cycles, optimization was conducted, and the usefulness of suggested modeling was confirmed.