Driving Cycle Construction And Variable-horizon Predictive Energy Management For Plug-in Hybrid Electric Bus

With great potential for industrialization and marketization,Plug-in Hybrid Electric Vehicle(PHEV)is very beneficial for the solving of air pollution,improving fuel economy and sustainable development of automobile industry.In this paper,Plug-in Hybrid Electric Bus(PHEB)as the main research object,and by the building of urban driving cycle of Zhengzhou,online reconstruction of future driving cycle and variable-horizon prediction,regional oriented online predictive energy management method are proposed.The main research achievements are as follows:(1)Driving cycle of Zhengzhou(ZZUDC)is constructed.Data acquisition test of PHEV’s original driving cycle in Zhengzhou is designed and carried out.To assess the accuracy of driving cycles,Performance Measurement(PM)indicators based on the driving cycle property parameters are determined.According to Markov stochastic process theory,Monte Carlo method and PM indicators,ZZUDC is constructed.The correlation coefficient between characteristic parameters of this constructed ZZUDC and original driving data reaches 99.72%,which has improved by 2.4% compared with traditional construction method.The construction of ZZUDC is the foundation of regional oriented online predictive energy management.(2)The selection result of the power system of Plug-in hybrid electric bus is verified.Based on the power density and velocity density of ZZUDC,the selection of the engine,tracion motor,ISG motor,power battery pack and single reduction ratio are verified being reasonable by weight coefficient method.An improved EV+CD+CS rule-based energy management strategy are proposed.The simulation taking ZZUDC as target velocity shows the selection result of PHEB power train satisfies the power demand of urban roads in Zhengzhou.The improved rule-based energy management strategy makes the switch between CD mode and CS mode possible and extends the working time of CD mode.One-hundred-kilometer fuel consumption and one-hundred-kilometer electricity consumption are 21.01 L and 13.52 kWh respectively,which proves that the dual motor coaxial hybrid configuration has better fuel economy.(3)The global optimization of PHEB’s energy management is completed.To reduce the calculation amount of DP algorithm,effective control sets of control variables are determined on the basis of working modes of PHEB.After DP global optimization,one-hundred-kilometer fuel consumption and one-hundred-kilometer electricity consumption are 16.9766 L and 11.6317 kWh respectively.These are the theoretical optimal consumption values,which offer as reference for the assessment of predictive energy management strategy.(4)Multi-scale Single-step model is built for the prediction of future vehicle driving conditions.According to the correlation coefficient of adjacent sample points in most driving cycles on the world,the maximum prediction interval is set as 35 s.After analyzing the Root Mean Square Error(RMSE)of prediction results with different prediction time intervals,it is confirmed that the accuracy of Multi-scale Single-step method is better than Fix-scale Multi-step method.To solve the shock and divergence of Multi-scale Single-step prediction results,four points mean filter and quadratic polynomial fitting are applied for optimization.Simulation results verify that this kind of optimization can solve these two problems effectively and the accuracy has been improved by more than 10%.(5)A predictive energy management strategy based on multi-scale variable time domain prediction model is proposed.The phenomenon that missing driving conditions exists in actual predictions due to the limited driving cycle data is explained,and then a kind of online reconstruction and state filling method are proposed.The principle of time domain transition is established and variable time domain real-time prediction model is built,which are introduced into the framework of Model Prediction Control(MPC)with DP algorithm.The results of Hardware-in-the-loop(HIL)simulation show that the prediction time domains spread within 35 s and the prediction accuracy is 8.203km/h,improved by 21.14 compared with fix time domain prediction.Compared with the optimal results of traditional fix time domain MPC,the one-hundred-kilometer fuel economy of variable time domain MPC is 18.3485 L reducing by 6.95% and the one-hundred-kilometer electricity consumption is 13.1081 k Wh reducing by 5.54%.