Research On Multi-objective Energy Optimization Of Plug-in Hybrid Electric Vehicles Based On Data

In this paper,for a parallel plug-in hybrid electric vehicle(PHEV)with an electronically controlled automatic mechanical transmission(AMT),three predictive models based on machine learning algorithms to predict vehicle speed in the short-term time domain are established.Model predictive control(MPC)algorithm is used as the energy management framework,and the Pontryagin minimum principle(PMP)is used as the solution algorithm,and the multi-objective energy management optimization considering the influence of battery temperature is carried out.The specific research content is as follows:First of all,based on the Autonomie database,this paper establishes a mathematical model of the important components of the PHEV power system: the quasi-static model of the engine,motor,battery,and transmission,as well as the longitudinal dynamics model of the vehicle,and the different work mode of the PHEV is analyzed.A rule-based energy management strategy is established.Second,based on machine learning algorithms,three short-term vehicle speed prediction models in the time domain are established,including Back Propagation Neural Network(BPNN),Support Vector Machine(SVM)and Extreme Learning Machine(ELM),the root mean square error(RMSE)of these three different predictor models is compared,and it is found that the prediction effect of BPNN and SVM is equivalent,and the prediction effect of ELM is slightly worse,but the prediction speed of ELM is faster.The influence of historical driving data of real vehicles on the prediction effect of the prediction model is explored,and it is found that compared with standard driving cycles data,using historical driving data of real vehicles as the training set can achieve better prediction results.Third,in order to compare the optimization effect of the MPC algorithm,global energy management strategies are established,including the Dynamic Programming(DP)algorithm and the Pontryagin minimum principle(PMP).The basic principles of DP and PMP are explained;for the shooting methods involved in PMP,the effects of three different shooting algorithms are explored,including dichotomy,genetic algorithm and particle swarm optimization algorithm,and it is found that all three shooting algorithms can meet the requirements and the dichotomy is faster to solve.At last,MPC energy management strategy based on PMP is established,considered power battery temperature model,a multi-objective optimization framework is established,and the influence of temperature weight coefficient on energy consumption and battery temperature is analyzed;combined with the short-term vehicle speed prediction model,Combined with the short-term vehicle speed prediction model,it is found that using the real historical data of vehicles after path screening as the training set can obtain better prediction results and better fuel economy.Compared with optimizationbased and rule-based energy management algorithms,the effectiveness and feasibility of the proposed algorithm are verified.