In order to solve the problem of low accuracy in predicting the remaining service life of proton exchange membrane fuel cells, this paper proposed a dynamic fuel cell Remaining Useful Life (RUL) prediction model based on Northern Goshawk Optimization (NGO), Convolutional Neural Network (CNN) and Bi-directional Long Short-Term Memory (BiLSTM) neutral network. Firstly, NGO optimized the learning rate, hidden nodes and regularization coefficient of the CNN-BiLSTM model, and then the CNN-BiLSTM model extracted the features of the input data through the convolutional layer, and input it into the BiLSTM layer for timing modeling and prediction. In addition, wavelet threshold de-noising algorithm was used to smoothen the original data. Pearson correlation coefficient was used to extract model input variables, and NGO-CNN-BiLSTM network power prediction model was built. The simulation and verification results show that this method can effectively improve the prediction accuracy of the remaining service life of fuel cells up to 99.49%, which is higher than that of other comparative models.

With a light truck equipped with power-following fuel cell as the research object, a power-following energy management strategy of a fuel cell based on wavelet transformation has been constructed combining the frequency characteristics of the vehicle power demand under typical driving cycles using wavelet analysis method. The low frequency component of the vehicle power demand obtained after wavelet transform was allocated to the fuel cell, so as to improve the fluctuation of the fuel cell output power. In addition, the wavelet decomposition level was dynamically adjusted to regulate the power following degree automatically, to ensure the power supply of the vehicle under various power battery charge-discharge capacity and vehicle power requirement. The simulation result shows that the power-following allocation strategy based on wavelet transform can meet the vehicle power demand under typical driving conditions, and the SOC can be kept in the reasonable range, the power output fluctuation of the fuel cell has been decreased by 29.8%, in addition, the hydrogen consumption has been reduced by 1.5%.

To realize the coordinated control of the flow pressure of the air subsystem for Proton Exchange Membrane Fuel Cell (PEMFC), this paper applied the fuzzy control principle and the Mamdani inference method to reason the control rules of fuzzy decoupling control, to design the fuzzy decoupling composite controller with MAP feedforward control. Finally, the comparison test with the fuzzy PID controller and the supplementary environment adaptation test were carried out. The results show that the fuzzy decoupling controller can achieve better control effect and has good environmental adaptability with flow control error less than ±3 g/s, and pressure control error less than ±10 kPa.

This article analyzed the technical requirements of hydrogen sensors used in fuel cell vehicles, and studied the working principles, characteristics of various vehicular hydrogen sensors, compared the performance and parameters of mainstream hydrogen concentration sensors at present, summarized and analyzed the current development status of hydrogen concentration sensors. The technical development trend of hydrogen concentration sensors was predicted based on industry development needs.

In the state-of-charge estimation of power battery, the traditional Extended Kalman Filter (EKF) ignores high-order terms and Particle Filter (PF) suffers from particle degradation and loss of diversity during the resampling process. To address this issue, this paper proposed the improved Mixed Kalman Particle Filter (MKPF) algorithm. Firstly, the extended Kalman filter was used to generate the state estimate of the system, and then the unscented Kalman filter was used to repeat the process. The state estimates obtained by the extended Kalman filter and the unscented Kalman filter were used together as the particle filter proposal distribution, and value sorting was used to determine the survival of the fittest particles. Simulation and experimental results show that the maximum error of SOC estimate by the proposed algorithm is 1.2%, which is better than the estimation accuracy of the existing PF, EKF, and UKF algorithms on SOC.

Current Time Of Flight (TOF) camera products have problems such as small viewing angle, and inability to consider all passengers in the front row. To address this issue, this article proposed a design method for TOF cameras, which solved the heat dissipation problem through thermal design simulation, and effectively expanded the camera’s Field Of View (FOV) through optical design and simulation of the TOF camera lens. At the same time, a complete design verification test was conducted to verify the performance of the TOF camera product. The results show that the system fully realizes the functions of the TOF camera, cover the passengers in the front row with a large FOV, and meets the performance requirements of the automotive grade.

To establish an accurate fuel consumption prediction model of heavy-duty diesel vehicles, this paper firstly used the dataset collected by heavy-duty diesel vehicles in real road driving, and Pearson correlation coefficient to calculate the correlation between different factors and fuel consumption, then selected 7 factors with strong correlation with fuel consumption, and used Back Propagation (BP) neural network and Long Short-Term Memory (LSTM) neural network to establish fuel consumption prediction models for heavy-duty diesel vehicles. The prediction results of different driving sections show that the prediction accuracy of BP neural network for fuel consumption values in different road sections differs sharply, and the generalization of the model is low, while the prediction of different road sections of the LSTM model is very accurate, and the model generalization is strong.