The energy management strategy of hybrid electric vehicle(HEV) directly determines the fuel economy, driving performance and life of vehicle. In order to solve the contradiction between the optimal energy management strategy of HEV and the uncertainty of real-time driving conditions, the energy management strategy of HEV based on model predictive control(MPC) and Dung beetle optimizer(DBO) was proposed based on the research object of hybrid HEV. First, the strategy used the vehicle speed prediction model to predict future driving speed based on stacked long-short term memory neural network(Stacked LSTM-NN). Then, according to the predicted vehicle speed, the power distribution problem of HEV was described as a rolling optimization problem within the MPC prediction range. Considering the cost function of fuel consumption and battery protection, the DBO algorithm was used to optimize the engine power in the forecast time domain. Finally, under urban dynamometer driving schedule(UDDS) conditions, the speed prediction accuracy and fuel economy of proposed strategy were simulated and compared with other strategies. Compared with the traditional LSTM speed prediction model, the RMSE of Stacked LSTM speed prediction model reduces by 13.9%, and the average prediction time of each step reduces by 1 ms. Compared with the rule-based strategy, the fuel saving rate of DBOMPC strategy model reached 25.3%, and the SOC state is more stable and the battery protection effect is better.