Self-learning Algorithm Of City Driving Cycles For Electrical Buses

Faced with the pressure of the energy and environment, more and more automobile companies begin to develop electric vehicles,including hybrid vehicles, pure electric passenger cars and buses. Optimization of almost all the electric vehicles'energy efficiency depends on the driving cycles, in fact, it's difficult to know a city's driving cycles during the vehicles'development and optimization before driving enough distance. So it's very important to develop a self-learning algorithm to get the driving cycles during running.Based on the data from the CAN bus data recorder of XL electric vehicles,developed by Tianjin University. In the paper , a SOM neural networks was developed to realize a learning algorithm of driving cycles.All data were divided a group of driving sequences,28 Eigen values were defined to describe every driving sequence, which were set up a driving sequence database. Through a multivariate statistical tool, the main factor was analyzed, and a SOM neural network was used to cluster and set up the driving cycles according a set of defined conditions.Three kinds of driving cycles, jammed, expedite and synthetical driving cycles were set up. Compared with other driving cycles, the driving cycles set up in this paper can meet the general regulation, and can describe the basic characteristics of road transportation situation in Tianjin. The results show the ability of the SOM neural networks'self-learning, especially which can cluster exactly new coming data. It also proves the possibility of driving cycle learning on line.