Research On Gearshift Control Optimization Of Electric Vehicle With Dual Clutch Transmission

Gearshift control problem for multi-speed transmissions of electric vehicles,especially the gearshift control optimization problem considering object nonlinearity and condition uncertainty in the actual shifting process,is one of the technical difficulties and research hotspots in the field of electric vehicle powertrain technology.In this paper,a pure electric vehicle equipped with a two-speed dual-clutch automatic transmission was taken as the specific object and the study of gearshift strategies and controller optimizaition for the actual gearshift process was carried out.We detaily analyzed the working principle of the dual-clutch automatic transmission and divided the gearshift process into torque phase and inertia phase.Considering the dynamic characteristics of the rotating components of each stage,we established an 11-degree-offreedom system model and a longtitude vehicle dynamic model,formulated Power and Economical shift schedules that are suitable for pure electric vehicles and simulated the shift under WLTC,which verified the correctness of the model and the gearshift schedule.A feedforward-feedback gearshift control model was established,in which the feedforward control was based on optimal trajectory of gearshift strategy.Based on the Legendre pseudospectral method,a multi-objective optimization of the gearshift strategy was carried out.Gearshift time,friction work and jerk were selected as evaluation indicators of the gearshift performance;motor torque and clutch transfer torque were considered as control variables.The Time-pattern and Jerk-pattern optimization trajectory were obtained by selecting different weight values of performance index.Simulation of PID feedback gearshift control for 11-degree-of-freedom vehicle systems was conducted,which we found that control stability and robustness were bad.Aiming at the optimization problem of gearshift control of object nonlinearity and uncertainty of working conditions during actual gearshift process,we used the Gaussian kernel radial basis function neural network RBFNN controller for feedback control,and the PILCO reinforcement learning algorithm was applied for controller parameter optimization.Gaussian process was used to learn the probability dynamic model of vehicle nonlinear system;the distribution mean of real-reference speed difference of clutch at the current step was input into the Gaussian kernel-radial basis function neural network;the difference distributions of speed differences for all steps and value function were calculated.After the multiple iterations of the PILCO reinforcement learning algorithm,control policy for actual gearshift control was improved trough optimizing conreoller parameters.A comparative analysis of the gearshift control optimization effect was carried out for different road slopes and vehicle loads.The results showed that the PILCO-optimizaed RBFNN controller has a good gearshift control effect during the actual gearshift process,and it has good adaptivity and robustness in complex driving conditions such as different slopes and loads.The research work in this paper has extensive reference signification for the control optimization of different types of multi-speed transmissions of electric vehicles under complex driving conditions and the calibration of controllers in engineering development.