The Research On Predictive Control Strategy Of Plug-in Hybrid Electric Vehicle

The plug-in hybrid electric vehicle combines the advantages of electric vehicles and hybird electric vehicles.It is an solution to reduce the energy consumption and emission of polluting gases.It is also a important transition products from internal combustion engine(ICE)to electric vehicles.The energy management strategy is a main factor to affecting vehicle fuel reducing result.It determines how the two power sources to work together to maximize the advantages of hybrid vehicle.Traditional PHEV adopt rule-based energy management strategies that cannot make decisions based on the current driving cycle and real-time vehicle information.The energy management strategies based on optimization theory can solve this problem well.In this paper,a GT-Suite and MATLAB/simulink co-simulation platform was established to build a full vehicle model and a rule-based energy management strategy based on the CDCS rule was implemented.The dynamic programming algorithm was used to calculate the global optimal solution for the specific driving cycle.The influence of the SOC discrete distance on the optimization results and calculation time were analyzed and compared.The dynamic programming was used in the model predictive control framework and the SOC reachable region was determined to reduce the calculated amount.The influence of the length of different prediction horizon on the optimization result of the current step was analyzed.By using polynomial fitting instantaneous fuel consumption of engine and the SOC gradient of battery with load torque and speed,the problem in prediction horizon was adjusted to a linear constraint optimization problem,which greatly reduced the calculation time and improved the real-time performance of algorithm.Based on the scenarios that the traveling distances are known and unknown,two types of SOC constraints were discussed to make the SOC value of the battery change more reasonably.In order to establish a more actual vehicle speed prediction model,the on-board diagnostic equipment was used to get the driving information of an actual driver as the training samples and test samples,which included high-speed,rural,and urban cycles.The artificial neural network was used to predict the vehicle speed and the relationship between the prediction accuracy with the length of input duration and different variables were analyzed and compared.The vehicle speed prediction model based on BP neural network and RBF neural network were established respectively.The key parameters were selected and determined based on empirical formulas and the initial value of neural network were optimized using particle swarm optimization algorithm.The forecasting result was improved compared with the prediction method with input of speed only.Through the simulation of a series of standard driving cycle,the simulation results of various control strategies in high SOC state and SOC maintenance state were compared and analyzed.The results prove that the method used by the author is effective in improving the fuel economy and improving the real-time performance of the algorithm.Finally,by comparing the simulation results under different vehicle speed prediction models,the influence of the accuracy of the prediction results on the actual fuel consumption is analyzed,and the necessity of the accurate vehicle speed forecasting model is confirmed.