Research On Parameter Matching And Energy Management Strategy Of Fuel Cell Commercial Vehicle Power System

The energy crisis and climate issues have become important issues related to the national economy and the people’s livelihood.Automobiles are one of the main sources of energy consumption and pollutant emissions.Fuel cell vehicles are attracting more and more attention because of their clean and pollution-free characteristics.Energy management directly affects the power,economy and durability of the vehicle.How to manage the distribution of multiple energy sources in the vehicle is the key to the power system integration technology of fuel cell vehicles.This paper takes the fuel cell commercial vehicle as the research object to carry out the research on the integrated development of the whole vehicle,focusing on solving the problem of the energy management of fuel cell vehicles.The main contents of this paper are as follows:Power parameter matching for fuel cell commercial vehicles: Firstly,numerical analysis was conducted to compare the performance of power systems with different architectures in terms of power and economy,and the most suitable power system architecture for commercial vehicles was selected.In terms of calculating the parameters of key components in the power system,a parameter matching method based on reverse derivation of design indicators was adopted,and the basic parameters of fuel cells,power batteries,and drive motors were preliminarily calculated based on the overall vehicle design and assessment indicators.Considering the evaluation indicators of system power,economy,and reliability,combined with hydrogen consumption model and integrated fuel cell service life factors,the influence of different degrees of mixing on maximum speed,usage cost,energy component cost,and component life decline was analyzed.At the same time,a multi-dimensional evaluation system was established based on the above indicators,and power system parameter matching and optimization design were carried out using mixing degree.Finally,the rationality of parameter matching was verified through the forward fitting parameter verification method based on virtual simulation,and the closed-loop design of vehicle parameter matching was completed.Dynamic system modeling: The establishment of dynamic system model is the basis of energy management simulation.In this paper,the non-linear dynamic model of components and the multidomain energy dissipation model of the vehicle were mainly established.In order to reflect the efficiency characteristics,dynamic output characteristics and durability characteristics of fuel cell,fuel cell efficiency model and electrochemical model were established.In order to reflect the durability of fuel cell,a long short-term memory artificial Neural Network(LSTM)was used to carry out the life prediction model and to build the durability model of fuel cell.The internal resistance model,thermodynamic model and durability model of the power battery were established to fit the input and output characteristics,heat loss and life degradation characteristics of the power battery.Drive motor model including efficiency model and mechanism model was established.At last,vehicle model including driver model and energy dissipation model was established.Vehicle model validation and power system integration: fuel cell energy-saving and highefficiency durability test bench and performance test bench were designed,which solved the serious energy waste during the durability test of fuel cell.Relevant key data were obtained through bench test,and fuel cell life prediction model and durability model were improved based on the durability test data.A safe test bench for drive motor was designed,which improves the test safety and verifies the accuracy of the drive motor model.3D digital models of the whole vehicle and its components were designed,and components were placed on the whole vehicle according to their size parameters.The communication protocol of fuel cell,battery and other components was established,and the communication standard for data interaction between different components was established.The integral electrical schematic diagram of the whole vehicle and its components was drawn,and the low-voltage and high-voltage wiring harnesses were designed according to the electrical schematic.The components and high-voltage wiring harnesses were arranged on the prototype vehicle according to the 3D model and electrical schematic.Fuel cell controller was designed and developed,and fault information analysis,data differentiation,preventive maintenance and faulttolerant control were performed based on test data.Finally,the accuracy of the developed vehicle pure electrical model and mixed model was verified by vehicle bench test and road test,and a complete model development and verification system was formed.The test results showed that the maximum error of 100 km power consumption under pure power mode was 3%.In the hybrid mode,the measured hydrogen consumption of 100 km was 1.68 kg,the simulated test hydrogen consumption of 100 km was 1.56 kg,and the error of simulation and test hydrogen consumption of100 km was 0.12 kg and the error percentage was 6.97%.Design and validation of energy management strategy considering durability: in order to maintain the SOC of power battery within the ideal range,thermostat control strategy was used as the external frame and the fuel cell operating range was divided by combining the SOC status of power battery;Secondary utility functions were used in the SOC maintenance interval to calculate the output power of fuel cells and power batteries.In order to solve the unknown parameters in the quadratic utility function,three solutions were proposed: the first one used model prediction algorithm to construct long/short-term vehicle operating condition prediction method based on multi-source information fusion,taking the minimum equivalent hydrogen consumption,the lowest performance degradation and the dynamic output characteristics of fuel cell as the indexes.LSTM was used to forecast the speed and PSO(particle swarm optimization)was used to optimize the network layers of LSTM,the number of neurons in each layer,the number of fully connected layers and the number of neurons in the fully connected layer.The demand power sequence within the forecasting horizon length was calculated according to the forecasting speed,and the demand power sequence within the forecasting horizon length was solved by using dynamic programming algorithm,and the unknown parameters in the utility function were deduced.The second solution was to solve Pareto solution of unknown parameters in utility function by means of iteration simulation optimization by using artificial bee colony algorithm and combining with multi-objective optimization algorithm.The third solution: the DDPG(deep deterministic policy gradient)algorithm was used to design the objective optimization function and used the model to learn iteratively to eventually obtain the unknown parameters in the utility function.Three solutions were compared in this paper.The results showed that MPC algorithm performs best in reducing life degradation and increasing driving mileage,MOABC(multi-objective artificial bee colony algorithm)performed worst,DDPG was also close to MPC in driving mileage and component life degradation.Finally,the accuracy of DDPG algorithm was verified by model in the loop and hardware in the loop.The maximum speed error of HIL test and simulation test was within 1m/s,the maximum speed error percentage was 0.71%,the maximum power error was 1.3k W and the maximum power error percentage was 2.2%.