Maximum power tracking-based adaptive cold start strategy for proton exchange membrane fuel cell

•Establishing a PEMFC output voltage evolution prediction model for cold start conditions.•Proposing an adaptive strategy to provide the best starting current load for PEMFC cold start.•Screening learning targets through a multi-attribute evaluation model based on fusion methods.•Analyzing the effect of starting temperature and starting current load on the evolution of PEMFC performance. The difficulty of cold start of proton exchange membrane fuel cells is one of the key factors limiting their commercialization. To determine the optimal cold-start load current to balance PEMFC icing and internal reaction heat generation, an adaptive cold-start strategy for fuel cells based on maximum power tracking is proposed in this paper. First, the semi-empirical model of output voltage is improved and the recursive least squares method with adaptive forgetting factor is used for model parameter identification. Subsequently, a multi-parameter evaluation model based on gray correlation analysis and technique for order preference by similarity to ideal solution is established for screening the optimal start-up time, and the corresponding load is used as the learning target of the output voltage prediction model. Finally, the output voltage model is used for maximum power point tracking, and the resulting current load is used for cold-start experiments of the PEMFC. The results of the study show that the PEMFC performs the best overall at 360 S under experimental conditions according to the evaluation model. The proposed strategy based on the new output voltage prediction model can reduce the prediction deviation of the maximum power point at 360 S by 40.64 % compared to the cold start strategy using the benchmark empirical model. The 27 A current load obtained by the proposed strategy results in a high and stable output performance of the PEMFC cold start, which verifies the feasibility of the strategy.