Abstract: A decoupling optimization algorithm with layered method was proposed. Multi-objective particle swarm optimization was used to optimize the parameters of vehicle's power components at outer layer. Meanwhile, efficiency optimizations of the battery and inverter were introduced to extract and transfer the different optimization fruits to inner layer. Bellman dynamic programming algorithm was used at inner layer. The optimal shifting strategy was established by solving the parameters of power component from outer layer optimizations. Based on this, the gear shifting optimization results of dynamic planning were extracted by GRNN, and an adaptive driver model and a vehicle forward simulation model were established by using the gear shifting strategy obtained. Aiming at the power and economy, the layered optimization results were further selected by the vehicle forward simulation analysis. The results show that the optimization algorithm realizes the decoupling of shifting control strategy and the parameters of the power components, effectively improves the efficiency of the optimization, and at the same time, the global optimization results may be obtained, which means that the vehicle economy improves significantly.

Key words: electric vehicle(EV), multi objective decoupling optimization;generalized regression neural network(GRNN)；adaptive driver

摘要： 提出了一种采用分层方法的解耦优化算法。外层采用多目标粒子群算法对整车动力部件参数进行优化，同时引入对电池以及逆变器的效率优化，将得到的不同优化结果实时提取并传递给内层；内层采用Bellman动态规划算法，根据外层优化得到的动力部件参数求解，建立优化后的换挡策略。在此基础上，通过广义回归神经网络提取动态规划的换挡优化结果，利用所得到的换挡策略建立了自适应驾驶员模型和整车正向仿真模型，以动力性和经济性为目标，通过整车正向仿真分析对分层优化结果进行进一步选择。研究结果表明，该优化算法实现了换挡控制策略与动力部件参数的解耦，有效提高了优化效率，同时能够获得全局优化结果，明显提高了整车经济性。

关键词: 纯电动汽车, 多目标解耦优化, 广义回归神经网络, 自适应驾驶员