Study On Urban Driving Cycle Construction And Its Multi-scale Prediction For Hybrid Electric Vehicles

In this thesis, the driving data of plug-in hybrid electric bus has been acquired in a data acquisition experiment; According to the statistical analysis based on these driving data, the driving characteristic of hybrid electric bus is defined, driving condition’s dividing principles and criteria of Performance Measure are determined; An urban driving cycle is constructed based on theories of Markov chain and characteristic parameters such as velocity, acceleration and road slope. Then the multi-scale prediction method is used to establish the driving condition prediction model based on the constructed urban driving cycle which is treated as historical information in the prediction model. To minimize the root mean square deviation of prediction results, Mean filter and Least squares method are used in prediction model. Finally, the prediction model is applied in real-time simulation system. The prediction methodology under the driving-cycle-missing-state is developed. The accuracy estimation method of vehicle driving condition in real-time prediction model and feasible prediction duration analyzing mechanism are proposed. The driving condition real-time prediction experiment is accomplished on hardware in loop(HIL) simulation platform.(1) Considering the differences between hybrid cars and traditional fuel vehicles, the partition principles of velocity fragments in driving conditions are formulated. The choosing and evaluation mechanism of parameters during the driving cycle construction, such as velocity threshold, acceleration threshold, PM(Performance Measure), etc is proposed. In order to merge the road gradient into driving cycle, the vehicle instantaneous output power and nominal velocity are introduced to distribute velocity fragments. Finally, theories of Markov chain are applied to construct the driving cycle for hybrid electric bus. Comparing with the driving cycle which is constructed based on traditional micro-trips method, the driving cycle constructed based on Markov method could be a better result.(2) According to the correlation analysis of adjacent velocities, the driving cycle Markov property is verified. Two methods are proposed to predict the future driving condition of hybrid electric bus. Because of the better performance in prediction accuracy, multi-scale prediction is chosen in this study. To minimize the root mean square deviation of prediction results, mean filter and least squares method are used in prediction model, and the integral driving condition prediction model is constructed.(3) The prediction model is applied in real-time simulation system. The prediction methodology under the driving-cycle-missing-state is developed. The accuracy estimation method of vehicle driving condition in real-time prediction model and feasible prediction duration analyzing mechanism are proposed. Driving condition hardware in loop(HIL) simulation platform is established based on Moto Tron & VT System, which could resolve the pedal signal into current velocity. The driving condition real-time prediction experiment is accomplished on this platform, and prediction accuracy and operation speed are tested.