混合动力汽车作为我国新能源战略的重要发展方向，如何优化整车能量分配以提升动力性和燃油经济性是研究热点。本研究针对插电式混合动力汽车（PHEV）的能量管理策略（EMS），在随机模型预测控制（SMPC）框架下，构建了基于卷积神经网络与长短期记忆网络（CNN-LSTM）的车速预测模型，并通过蜜獾算法（HBA）优化其隐藏单元数、最大训练周期和初始学习率。试验表明，HBA-CNN-LSTM模型的各评价指标均优于对照方法。随后，以PHEV动力系统为研究对象建立仿真模型，将HBA-CNN-LSTM与SMPC结合，利用动态规划（DP）以发动机油耗最小为目标进行优化。仿真结果显示，在5个用户工况下，所提出策略相较于CDCS策略油耗降低0.240 8 L（10.491%），与DP策略的差距仅为0.087 3 L（3.802%）。

As an important direction of China's new energy strategy, optimizing the energy allocation of hybrid vehicles to improve power and fuel economy is a research hotspot. For the energy management strategy of plug-in hybrid electric vehicles (PHEVs), under the framework of stochastic model predictive control (SMPC), a vehicle speed prediction model based on convolutional neural network and long short term memory network (CNN-LSTM) was proposed, and the number of hidden units, maximum training period, and initial learning rate were optimized using honey badger algorithm (HBA). The experiments showed that HBA-CNN-LSTM was superior to the control method in all evaluation indicators. Subsequently, the simulation model of PHEV power system was established. The HBA-CNN-LSTM was combined with SMPC, and dynamic programming (DP) was used to optimize according to the goal of minimizing engine fuel consumption. The simulation results show that under 5 user-defined operating conditions, the proposed strategy reduces fuel consumption by 0.240 8 L (10.491%) compared to the CD-CS strategy, and the difference with the DP strategy is only 0.087 3 L (3.802%).