Abstract A novel control strategy for the adaptive real-time energy management of a commuter pull-in hybrid vehicle is proposed. The proposed strategy can adapt to various driving conditions so that fuel economy can be improved further in practice. Its main feature is that a fuzzy inference system (FIS) for online estimation of the reference SOC and an adaptive update law with traffic recognition are blended into the main frame of an adaptive-equivalent consumption minimization strategy (A-ECMS). The FIS is established through an adaptive neuro-fuzzy inference system (ANFIS) that is offline trained by the traffic information extracted from historical traffic data and the reference state of charge (SOC) optimized by dynamic programming (DP). The adaptive update law with traffic recognition means that the adaptive equivalent factor (A-EF) of the real-time A-ECMS is adjusted online according to the traffic information in the real route besides the SOC of the vehicle battery. This is because the initial A-EF and the proportional–integral coefficients of the A-EF adjuster are mappings of the SOC and the traffic road segment, and the mappings are optimized by particle swarm optimization (PSO) according to the different initial SOC and the real historical driving cycles of each segment. The proposed strategy is carried out on the simulation test platform integrated GT-Suite simulator and MATLAB/Simulink. The simulation results show that the proposed strategy can reach an optimal energy distribution on a near global optimal level (close to the level of dynamic programming (DP) under the deterministic driving condition). Compared with a rule-based (RB) strategy, the traditional ECMS, an A-ECMS with the linear SOC reference, an A-ECMS with the EF optimized by PSO and an A-ECMS with the A-EF adjusted by a fixed PI feedback controller of the SOC, the fuel consumption is reduced by an average of 22.98% 10.26% 6.52% 2.33% and 5.91% respectively.

中文翻译：

---

基于模糊推理系统的插电式混合动力汽车自适应实时能量管理控制策略

摘要 提出了一种用于通勤混合动力汽车自适应实时能量管理的新控制策略。所提出的策略可以适应各种驾驶条件，从而在实践中进一步提高燃油经济性。其主要特点是将用于在线估计参考 SOC 的模糊推理系统 (FIS) 和具有交通识别的自适应更新律融合到自适应等效消耗最小化策略 (A-ECMS) 的主要框架中。FIS是通过自适应神经模糊推理系统（ANFIS）建立的，该系统通过从历史交通数据中提取的交通信息和动态规划（DP）优化的参考荷电状态（SOC）进行离线训练。具有交通识别的自适应更新规律是指实时A-ECMS的自适应等效因子（A-EF），除了车辆电池的SOC外，还根据实际路线中的交通信息在线调整。这是因为初始 A-EF 和 A-EF 调整器的比例积分系数是 SOC 和交通路段的映射，并且根据不同的初始 SOC 和粒子群优化（PSO）对映射进行了优化。每个细分市场的真实历史驾驶周期。所提出的策略在集成GT-Suite模拟器和MATLAB/Simulink的仿真测试平台上进行。仿真结果表明，所提出的策略可以在接近全局最优水平（接近确定性驾驶条件下的动态规划（DP）水平）上达到最优能量分布。与基于规则（RB）策略相比，传统ECMS、具有线性SOC参考的A-ECMS、具有PSO优化的EF的A-ECMS和具有固定PI调整的A-EF的A-ECMS SOC的反馈控制器，油耗平均分别降低了22.98%、10.26%、6.52%、2.33%和5.91%。