EVT Hybrid Powertrain System Modeling Based On Graph Theory And Designs Screening Under Multiple Characteristic Driving Cycles

Hybrid electric vehicles are highly praised for their long-lasting mileage of vehicles with conventional internal combustion engines and low pollution of pure electric vehicles.As one of the key technologies,the research and development of hybrid electric powertrains have been a widespread concern.Electric Variable Transmission(EVT),which uses planetary gearsets as power coupling mechanism,has been extensively studied due to its compact structure and high efficiency,and has produced a large amount of research results.At present,combining theories of graphics and mathematical,most of the design researches on EVT hybrid system adopt the ways of exhaustive or enumeration schemes,which are inefficient and incomplete.Based on the research of the existing EVT hybrid system graph theory model,this paper introduces the machine learning algorithms and integrates the driving cycles information to carry out the modeling and scheme optimization of the single planetary EVT hybrid system under different characteristic driving cycles.The specific research content is as follows:(1)Based on the existing EVT hybrid system graph theory hierarchical graph model and adjacency matrix model,the key elements are extracted to form the EVT hybrid system graph theory matrix model.The EVT dynamics modeling process is introduced into the graph theory hierarchical drawing model,and the dynamics model of the graph theory hierarchical drawing model is established.(2)Fuzzy C-means clustering optimized by Genetic Algorithm(GA)and Simulated Annealing(SA)is applied to 10 standard driving cycles.Cluster analysis is used to extract three types of characteristic driving cycles,and the characteristic parameters of three types of characteristic driving cycles are calculated.(3)The mathmatical models of system components including vehicle,engine,motors and battery are established.The dynamic programming(DP)is used to simulate the scheme in the sample space to obtain the fuel consumption under multi-character driving cycles,and the fuel consumption sample space under multi-character driving cycles of EVT hybrid system is obtained.(4)The graph theory model extrated from the adjaceny matrix of the EVT hybrid system and the characteristic driving cycles' parameters are both normalizing and joined together as the input of the nueral network to find the best-matched parameters.The BP,RBF and GRNN neural networks are used to build the fuel consumption models of the EVT hybrid system under different characteristic driving cycles.The test results are compared and the GRNN neural network fuel consumption model whose accuracy is the highest is the final selection.(5)The isomorphic analysis of EVT schemes in the sample space is carried out,and the fuel consumption model training samples are amplified by the isomorphic schemes.Besides the fuel consumption model is retrained by the amplified samples.(6)The Genetic Algorithm(GA)is used to search for best the EVT hybrid system schemes.At the same time,in the searching process,the screening conditions for the infeasible schemes are added,which is helpful for removing the infeasible schemes and ensures the efficiency of the scheme searching and the accuracy of the neural network fuel consumption model.