Search Results

Fuel cell reactors, as the core components of fuel cell vehicles, have a short life problem that has always limited the development of fuel cell vehicles. The life attenuation curve of fuel cell shows nonlinear characteristics, and there is no model that can accurately predict its effect. This paper is based on the experimental data of the vehicle fuel cell reactor, which is derived from the 600 h durability test run by a 4 kW fuel cell reactor. The relevance vector machine, as a Bayes processing method that supports vector machine, is a data-driven method based on kernel functions. The regression model is established by the relevance vector machine, and the super-parameters are found by genetic algorithm, because the kernel function strongly affects the nonlinearity of the curve, and the decay curve of fuel cell reactor performance is predicted according to four different kernel functions.

Fuel cell vehicle commercialization and mass production are challenged by the durability of fuel cells. In order to research the durability of fuel cell stack, it is necessary to carry out the related durability test. The performance prediction of fuel cell stack can be based on a short time durability test result to accurately predict the performance of the fuel cell stack, so it can ensure the timeliness of the test results and reduce the cost of test. In this paper, genetic algorithm-BP neural network (GA-BPNN) is proposed to modeling automotive fuel cell stack to predict the performance of it. Based on the strong global searching ability of genetic algorithm, the initial weights and threshold selection of neural networks are optimized to solve the shortcoming that the random selection of the initial weights and thresholds of BP neural network which can easily lead to the local optimal value.

In recent years, fuel cell vehicles have attracted more attention since the advantages of no environmental pollution and high energy density, however, the cost and durability of fuel cells have been important factors limiting the rapid development of fuel cell vehicles. How to quickly predict the life of fuel cells has always been the emphasis and focus of the industry. Therefore, this paper mainly focuses on two sets of proton exchange membrane hydrogen fuel cell durability test data. In this paper, we establish a fuel cell life prediction model to carry out product prediction research, using Gated Recurrent Unit Neural Network (GRU-NN)—a variant of “Recurrent Neural Networks” (RNN). This article first divides the two sets of fuel cell durability test data into a training group and a verification group and trains the established neural network model with the test data of the training group.

Fuel cell vehicles have always garnered a lot of attention in terms of energy utilization and environmental protection. In the analysis of fuel cell performance, there are usually some outliers present in the raw experimental data that can significantly affect the data analysis results. Therefore, data cleaning work is necessary to remove these outliers. The polarization curve is a crucial tool for describing the basic characteristics of fuel cells, typically described by semi-empirical formulas. The parameters in these semi-empirical formulas are fitted using the raw experimental data, so how to quickly and effectively automatically identify and remove data outliers is a crucial step in the process of fitting polarization curve parameters. This article explores data-cleaning methods based on the Local Outlier Factor (LOF) algorithm and the Isolation Forest algorithm to remove data outliers.