Predictive Control Of Energy Management For Plug-in Hybrid Connected Vehicle Based On The Road Condition Information

With environmental pollution and energy consumption becoming more and more serious,plug-in hybrid vehicle(PHEV)which is the transition product from traditional vehicle to pure electric vehicle has received extensive and continuous research.PHEV has effective energy-saving and emission-reduction performance.Intelligent and network connected are the development trend of future vehicles.The rapid development of the connected vehicles provides new ideas for development of energy management of PHEV.In this paper,the single shaft parallel plug-in hybrid electric vehicle is taken as the research object.The energy management of PHEV is developed based on the connected information.And the connected information contains vehicle location information from Global Positioning System(GPS),road slope information from Geographic Information System(GIS)and traffic information from Intelligent Transport System(ITS).The detail work is shown as follows:(1)Combining connected information with instantaneous optimization method,Equivalent Consumption Minimization Strategy(ECMS).Firstly,an equivalent factor estimation method based on demand energy per unit distance is proposed considering that PHEV can get energy form electric grid;then particle swarm optimization is used to optimized the method;finally,a neural network is used to predict the demand energy per unit distance,and ECMS based on prediction of demand energy per unti distance is developed.Then the proposed strategy is simulated under a real cycle to validate the preformance of proposed strategy.(2)Combining connected information with Model Predictive Control(MPC).Firstly,state space is established by multiple linear regression model of PHEV and a linear model predictive controller is developed based on the established state space;then connected information is used to predict future velocity and future reference SOC to improve the performance of MPC and hierarchical stochastic MPC is proposed.The proposed strategy is simulated under comprehensive cycle to validate the performance.(3)To develop universal eneryg management strategy,combining reference SOC eneryg management strategy based on Reinforcement Learning(RL)is established.And to improve adaptation of the RL based strategy,the possibility translation matrix of RL is update online.The proposed strategy is simulated under comprehensive cycle,Charging Depleting-Charging Sustaining(CDCS)and traditional optimization state simulation results are as comparisons.