Research On Scheme Design And Parameter Optimization Of EVT Hybrid Powertrain Based On Graph Theory And Machine Learning

With the acceleration of the industrialization process in various countries around the world,the per capita use of vehicles has risen rapiddly and the accompanying problems of environmental pollution and energy shortage have been getting worse.The hybrid vehicle adopts an energy-efficient driving method,which is small in pollution and long in cruising range.As a typical representative of hybrid powertrain system,EVT(Electrical Variable Transmission)system has broad development space and important scientific research value.In this paper,EVT hybrid powertrain is taken as the research object,and the following work is carried out by using the research methods such as graph theory and machine learning:(1)The hierarchical drawing model and matrix model of EVT system are established to improve the connection between multi-power source and planetary gear train components,and to enrich the research theory of graph theory in EVT system design.A graph theory dynamics model of the EVT system is established to analyze the interaction relationship between the components;(2)Based on 24 EVT basic schemes,a certain number of clutches/brakes are added to create a sample design space.Taking the configuration matrix that characterizes the topological relationship and performance characteristics of the EVT system as a sample input,the fuel economy of the EVT system is taken as a sample output.Using RBF and DBN network to learn the mapping relationship between EVT system configuration matrix and DP simulation fuel consumption,and to establish a configuration fuel consumption evaluation model,DBN configuration fuel consumption evaluation model performance is better by comparing and analyzing;(3)Based on the BPSO optimization method,combined with the DBN configuration fuel consumption evaluation model,a selective evolution strategy is added to enhance the convergence of BPSO,and the validity of the EVT scheme is judged.The search scope of BPSO is limited to the scheme design space to guarantee the accuracy and reliability of the configuration fuel consumption evaluation model to find the global optimal scheme;(4)Taking the optimal scheme obtained by BPSO optimization as the optimization object,four representative key parameters are selected,and Latin hypercube sampling is used in the set range to establish a parameter sample set.Using the ELM and GRNN network to establish the parameter fuel consumption model,GRNN parameter fuel consumption model performance is better by comparing and analyzing.Combined with PSO optimization,the optimal parameter combination is found.Therefore,the optimal scheme of EVT hybrid powertrain with optimal configuration and parameters is obtained.