A least-squares support vector machine method for modeling transient voltage in polymer electrolyte fuel cells

An investigation into the credibility and suitability of a transient voltage model that characterizes the dynamic behavior of polymer electrolyte fuel cells was carried out by means of quantitative and qualitative validations. The least squares support vector machine method was then used to construct a transient voltage model of a fuel cell in the first phase, including a validation based on experimental data obtained from a test rig. In the second phase, a thorough discussion of the effect of the fuel cell's operating conditions and the exterior load changes on the model's performance was implemented. For this phase, the influences of the sampling interval and ramp ratio are discussed and determined following a large number of tests under a variety of operating conditions. The results show that sampling with short time intervals is an effective way to improve the model's performance, and a smoother change to the exterior load is more likely to be approximated by the least squares support vector machine model. Moreover, the voltage model is sensitive to the ramp value by comparison to the ramp time. Suggestions for future applications of the transient voltage models are also provided. For a given combination of load changes, the sampling interval should be managed within a range to reach the demand data that satisfies the voltage accuracy. On the other hand, for a determinate sampling interval, the dynamic change of the load should be restricted within a limit to ensure that the model error is lower than the demand value.