Study On Energy Optimal Management For Series Hybrid Electric Vehicles

As the energy crises and environmental degradation are getting more and more remarkable in the world, researches of hybrid electric vehicles(HEV) are progressing gradually. And this dissertation studies energy management strategy(EMS) for HEV. The main work of this paper is described as follows:Based on empirical modeling approach with the aid of theoretical modeling, a forward model is presented which provides the necessary simulation platform for the development of EMS. Furthermore, this paper gives a systematical and comprehensive analyses to the working principle of series HEV's control strategy(CS), and thus the problems of series HEV's EMS to be resolved are ascertained, that is how to split the instantaneous power required between the two energy sources in order to minimize fuel consumption while the vehicle performs a given driving cycle, at the same time, the state of charge(SOC)of battery is balanced. The aim of the paper is to propose a real-time EMS which closer to global optimation, first combine theories of dynamic programming (DP) and monkey algorithm(MA), we get the control rules of global optimization, the rules are selected as pretreatment samples, then BP neural network based on fuzzy C-means clustering is proposed.Based on models of all parts and the problems to be solved by EMS, an optimized mathematical model of HEV is established according to the efficiency curve of engine and battery. For a given driving cycle, optimization problem of EM is a multi-stage decision problem, which can be converted to single-stage decision by discretization. Mechanism of minimization the fuel consumption in a given driving cycle is analyzed. Combining the characteristic of the problem and the theory of DP, improved DP(IDP) is presented. Simulation results show that IDP not only can get global optimal solution,but also can greatly reduce the computing time.Based on discretization mathematical model of whole driving cycle, a new intelligent algorithm, MA, is applied for solving the optimization problem. Output power of engine is used as a decision variable; balance of SOC is used as constraint equation. Chaotic-search, cooperation process and stochastic disturbance are added in original algorithms for fast convergence. The simulation results show that MA can get global optimal solution, and is effective to solve EM optimization problem of series HEV.Through analysis and simulation of IDP, EMS is abstracted as a nonlinear mapping of two inputs and one output, then BP neural network is designed for real-time EMS on the basis of fuzzy C-means clustering. The global optimum algorithm is adopted in several typical driving cycles, and the principles are taken as training samples for BP neural network. Training samples are classified by fuzzy C-means clustering algorithm, and then typical samples in each class are used for training neural networks off line, thus BP controllers corresponding to each class can be obtained. For the real-time data transmitted into the network, its distance to each cluster center is calculated respectively, and the BP network corresponding to the cluster center with minimal distance to the real-time data is selected, and the output of this BP network is used as the output of EMS. Simulation results show that BP neural network EMS based on fuzzy C-means clustering algorithm not only can simulate the global optimum principles, which can ensure a good fuel economy of HEV, but also can realize real-time control.