Research On Equivalent Consumption Minimization Strategy Of Hybrid Electric Vehicle Based On Parallel Computing

With the continuous growth of automobile production in China,the problems of environmental pollution and shortage of oil resources have become more and more prominent.In such an environment,Hybrid Electric Vehicle(HEV)came into being.It not only has the advantages of long battery life of pure fuel vehicles,but also has the characteristics of low emission of pure electric vehicles,so it has a good development and research prospect.Equivalent Consumption Minimization Strategy(ECMS),as a typical instantaneous optimization HEV energy management strategy,has great limitations due to its poor robustness.However,the classical adaptive ECMS(A-ECMS)strategy mostly uses Proportional Integral(PI)control,and its control performance is limited by the reasonable adjustment of the PI controller parameters,and requires repeated calibration to achieve better performance.Excellent,the control performance needs to be improved.With more and more scenarios where artificial intelligence technology is combined with industrial control,at the same time,since Deep Reinforcement Learning(DRL)has better self-learning and adaptive characteristics,many researchers will apply it to their respective Research areas.In this paper,DRL algorithm will be used to improve ECMS to form a new A-ECMS strategy.At present,the classical DRL algorithm needs to interact with the environment to collect a large amount of empirical data to gradually learn the optimal strategy,which results in slow algorithm iteration and low computational efficiency.In the context of the continuous development of "parallel computing + machine learning" technology,this paper proposes a parallel DRL-based algorithm for parallel HEV energy management to improve the iterative training speed of the algorithm.The main research contents of this paper are as follows:(1)The research status of HEV energy management strategies is expounded,and the characteristics,advantages and disadvantages of various strategies are introduced.(2)This paper presents the configuration of a parallel HEV,including its operating mode and energy flow.Then this paper models the various power components and traditional systems of the parallel HEV,and builds the vehicle model.Finally,this paper designs a control strategy based on deterministic rules,and verifies the HEV model.(3)Based on ECMS related theories,an ECMS control strategy model is built,and its control performance is tested under different conditions to verify the limitations of the algorithm.The control principle of A-ECMS based on PI is expounded,and then the algorithm model will be built in this paper,and then its control performance under different conditions and working conditions will be tested.(4)In this paper,a DRL algorithm will be designed based on the principle of parallel computing,and then a parallel DRL algorithm is designed according to the principle of parallel computing,and then the algorithm is built and deployed for experiments.Finally,a comparative analysis is made with the PI control-based A-ECMS strategy and the rule-based strategy under various working conditions..Through experiments,it is found that the parallel DRL A-ECMS strategy can effectively improve the training speed,and the control performance is better than the PI-based A-ECMS and rule-based strategies.