Powertrain Parameters Matching And Research For Electric Bus Based On Intelligent Algorithm

With the development and expansion of science technology constantly, experts and researchers from all fields are trying to combine knowledge from different subjects and using them in other subjects. One of these methods is intelligent algorithm, which origins from biology and is established through computer. Now it is used widely in different fields and it receives more and more approval and gets rapid development because of its advantages such as high accuracy and high efficiency. As a result, it is vital to have a further study on intelligent algorithm and its applications.Artificial Fish Swarm Algorithm (AFSA) and Support Vector Machine (SVM) are improved respectively and then used to get the most appropriate parameters of transmission ratios of electric bus. SVM are used to predict dynamic performance of electric bus, as well as economic performance, and then AFSA are used to optimize them. By doing these, not only the power consumption per hundred kilometers of batteries can be reduced but also the dynamic performance can be improved efficiently. Through this, the improvement of AFSA and SVM is testified and has significant.First in this paper, Genetic algorithm (GA), BP neural network, AFSA are introduced simply to have a certain understanding about them, as well as their advantages and disadvantages. Based on these preparation, do some improvement on AFSA in order to improve its diversity and balance the abilities of global searching and local approximation. To prove the efficiency of improvement, two typical testing functions are used to compare improved AFSA with other algorithms and the result is ideal.Then this paper also introduces SVM algorithm simply. In order to improve the performance of it, improvement of choosing basic parameters of SVM model is done by using improved AFSA. To proof the feasibility of improvement, GA, PSO, AFSA are used to choose SVM model’s basic parameters, too. Using three groups of data of classification from UCI standard database to prepare the results of these algorithm, it is verified that SVM combined with improved AFSA has a better regression performance.Next, an electric bus is transformed from a traditional fuel bus. Parameters of Electric motor, batteries and transmission ratio are matched preliminarily based on the requirements of electric vehicle performances and the characters of bus and an automatic transmission control strategy is established. The simulation model of electric bus is established through the software called AVL-Cruise. Transformation is proved to be accurate over basic performances by simulation.Last, on the basis of the electric bus model and according to different city bus driving road, both of dynamic performance data and economic performance data are gained by DOE plan, which are used to train SVM models. SVM models can use to predict dynamic performance and economic performance after trained. According to different city bus driving road spectrum and establishing multi-objective functions and optimizing them by using improved multi-AFSA, the optimum combination of transmission ratios can be got and verified by simulation.