Research On Energy Management Strategy Of Plug In Hybrid Electric Vehicle Based On Macroscopic Traffic State Perception

In the ongoing process of promoting green and low-carbon changes in the way energy is produced and consumed,the development of hybrid vehicles is an important means of reducing the level of end-use energy decarbonization.In order to fully utilize the energy saving potential of hybrid vehicles,it is important to develop an effective energy management strategy to rationally allocate the required power to individual power components.With the development of traffic prediction technology in Intelligent Transportation(ITS),accurate future traffic information provides a powerful guide for the vehicle energy management strategy.How to combine traffic prediction techniques with vehicle energy management strategies,improve the accuracy of traffic prediction methods as much as possible,and design hybrid vehicle energy management strategies that effectively utilize the perceived traffic state is the key to fully exploit the energy saving performance of vehicles.Since the macro traffic state can reflect the driving characteristics of bicycle,this paper conducts a research on the energy management strategy based on macro traffic state perception for the problem of how to use macro traffic information to guide the energy management control of bicycle.On the other hand,considering the complex nonlinear characteristics of traffic data in spatial and temporal dimensions,it is beneficial to improve the accuracy of traffic prediction by effectively capturing spatio-temporal correlation information,so the ability of traffic prediction models to capture spatio-temporal correlation of traffic data is investigated in this paper.Focusing on three key issues: the construction of training data for deep learning models,the construction of traffic prediction models,and the utilization of traffic information for energy management strategies,this paper mainly adopts the following solutions: constructing multi-dimensional traffic graph network data based on graph theory,then combining multi-graph convolutional neural network with multi-head selfattentive mechanism to construct a two-layer model to perform efficiently in traffic prediction tasks,and finally designing an equivalent fuel consumption minimization strategy(ECMS)combining traffic states predicted by traffic state perceptron to improve the fuel economy of vehicles.Firstly,an economic simulation platform for a P2 configuration parallel plug-in hybrid electric vehicle(PHEV)is constructed based on the Matlab/Simulink platform.The overall configuration,operating mode and basic parameters of the target model are introduced,and then the forward simulation modeling method is used to construct the vehicle model,driver model and control strategy model,in which the deterministic rule-based energy-linked management control strategy is introduced in detail and used as the benchmark strategy for vehicle energy efficiency evaluation.Secondly,the floating vehicle data obtained based on traffic big data is introduced,and the city is divided into different sub-regions according to the coordinate range of floating vehicles.Based on the statistical regional statistic features,a self-organizing mapping neural network(SOM)is used to classify the floating vehicle trajectory points into five grades,which are used to represent the current traffic state that the vehicles are in.The pre-processing work is performed for the floating vehicle data in each state,which mainly includes data point-intime synchronization,filtering and screening.Based on the processed trajectory data,Markov chain method is used to generate representative driving conditions for each class of data,and the consistency of the generated conditions with the original data is ensured through the verification of the representative conditions.Thirdly,the traffic prediction model dataset based on floating vehicle data is constructed based on deep learning methods,and considering the existence of spatio-temporal correlation characteristics of traffic data,the traffic graph network is constructed from multiple perspectives based on graph theory principles,and historical traffic data are used as model inputs.A two-layer model based on multi-graph convolutional neural network and multiheaded self-attentive mechanism is built to extract traffic data features from two aspects of spatial and temporal correlation,respectively.The constructed dataset is fully utilized to train the prediction model and to demonstrate the effect of training and validation.Then,the principle of the equivalence consumption minimization strategy is fully explained,and the importance of the equivalence factor in the strategy is clarified.An equivalent fuel consumption minimum energy management strategy based on traffic environment awareness is constructed and the optimization problem of the equivalent factor is solved using the Harris hawk optimization algorithm(HHO).Finally,the effect of the solution is shown based on the key problems out of this problem,i.e.,the test effect of the traffic environment perception model and the simulation results of the energy consumption simulation of the energy management strategy.During the testing of the environment-aware model,the overall statistical performance and the specific prediction effects are demonstrated,and the results show that the accuracy of traffic state prediction is as high as 90% and the deviation of grade prediction is controlled within about one grade;for the energy consumption simulation part,two real-world driving cycles are randomly selected for simulation in this paper,and the results indicate that using an ECMS strategy that incorporates macroscopic traffic states can improve the fuel economy of PHEVs by an average of 6.1% compared to a rule-based(RB)energy management strategy and a conventional ECMS energy management strategy.