基于行驶特征预测的HEV在线控制策略研究

Research on online control strategy of HEV based on driving features prediction

现有控制策略未考虑实际行驶特征对油耗的影响,离线全局最优策略虽能获得理论最小油耗,但只能离线应用。文章提出了一种基于行驶特征预测的控制策略,实现了离线最优轨迹的在线运用。采用动态规划(dynamic programming,DP)算法获得了离线最优轨迹;研究了汽车的行驶特征(包括行驶工况和行驶模式),利用欧几里得贴近度实现了行驶工况的预测;利用反向传播(back propagation,BP)神经网络进行了行驶模式预测;采用BP神经网络对离线最优轨迹及相应的汽车状态进行学习,设计了基于行驶特征预测的在线控制策略。仿真结果表明,与基于规则的电机辅助控制策略相比,该文设计的控制策略燃油经济性提高了7.51%,且工况适应性良好,同时电池电量保持能力较好。

The impact of actual driving features on fuel consumption is not considered into the existing control strategies. The offline global optimization strategy can obtain the theoretical minimum fuel consumption, but it can only be applied to offline state. So, an online control strategy based on the driving features prediction was proposed, and the offline optimal strategy was applied to the online control strategy. Firstly, the offline global optimal control trajectory was obtained by dynamic programming(DP) algorithm. Secondly, the driving features were studied, including the driving cycle and driving mode. Euclid closeness was adopted to predict the type of driving cycle. Moreover, the back propagation(BP) neural network was applied to predicting the driving mode. Then, the neural network studied the offline optimal trajectory and the vehicle states under each standard cycle, so an online control strategy based on the driving features prediction was designed. The simulation results demonstrate that compared with the rule-based motor assist control strategy, the designed strategy can be more efficient and the improvement of fuel economy is up to 7.51% and has good adaptability of cycle and battery capacity retention performance.