Data Driven Based Energy Management Strategies For Fuel Cell Hybrid Electric Vehicles

Fuel cell hybrid vehicles use fuel cells and batteries to form a hybrid system,which has the advantages of no pollution,short energy replenishment time and long mileage,and is considered to be an important solution to achieve carbon neutrality,so Governments and businesses globally have held it in high esteem.Due to the existence of various power sources,the energy management strategy is one of the key technologies to make the system work reliably,stably and efficiently.However,existing energy management strategies have problems such as short service life of power sources,poor adaptability to variable working conditions,and inability to be applied online.Aiming at the above problems,this paper establishes a set of data-driven energy management strategies by combining the advantages of traditional global optimization strategies and data-driven methods to improve the multi-objective optimization performance of the system and achieve a balance between optimality and real-time performance and improve its adaptability.The main research contents are as follows:(1)An in-depth examination of the composition and operational principle of fuel cell hybrid power systems,along with seven distinct energy management techniques,is conducted;At the same time,a comprehensive examination of the issues in the present energy management plan is conducted,with both the conventional energy management approach and the data-driven technique being scrutinized to furnish a basis for the design of the energy management plan.(2)An introduction of a fuel cell hybrid power system is made.The power system model is expounded in detail,the vehicle dynamics model of FCHEV is analyzed,and the fuel cell and storage battery models are established respectively.Then,the vehicle control strategy is built,and the finite state machine-PID model is used for power distribution,which verifies the correctness of the vehicle model and lays the foundation for the following text.(3)Aiming at the balance between real-time performance and optimality of hybrid power system,a real-time energy management strategy based on fuzzy C-means clustering and extreme gradient boosting algorithm is proposed.This strategy uses fuzzy C-means to cluster the working data set to reduce the subsequent optimization performance loss;Then use dynamic programming to solve the same type of working conditions to obtain the global optimal reference trajectory data set,and use the extreme gradient boosting algorithm to mine the deep feature information in the data set.Experimental results show that the strategy is very close to the global optimization effect of dynamic programming,and the calculation time is only 1/6 of it.It shows that this strategy can reasonably distribute power under the premise of real-time application,so as to achieve the purpose of improving fuel economy.(4)Aiming at the low adaptability of the hybrid system under variable operating conditions and the short service life of the power source,an adaptive energy management strategy considering multi-objective optimization is proposed.An extreme learning machine is employed by the strategy to detect the driving conditions in real time,thereby enabling the dynamic alteration of energy management strategy parameters.At the same time,a multi-objective optimization function considering both economy and durability is proposed to quantify the performance degradation of each power source more comprehensively and reliably.Experimental results show that this strategy can significantly reduce various unfavorable factors for power source attenuation,and achieve the effect of both economy and durability.At the same time,it can still maintain good optimization performance under complex working conditions,and has strong adaptability.