基于轨迹规划与CNN-LSTM预测的履带式混合动力无人平台能量管理优化

混合动力能量管理策略是混合动力系统的关键技术之一，对整车效率和燃油经济性等综合性能起到决定性作用。对于履带式混合动力无人车辆，其复杂的行驶工况对能量管理策略提出更高要求。在传统工况预测方法的基础上，提出一种基于无人驾驶轨迹规划的卷积神经网络与长短期记忆网络的预测模型。针对系统状态变量与控制变量搜索范围广、计算量大的问题，优化动态规划算法以获得最优控制序列；设计模型预测控制方法实现能量管理优化控制；通过实车试验进行验证。研究结果表明，采用卷积神经网络与长短期记忆网络的预测模型比基于规划速度的直接预测模型的精度提高了3%；基于该模型预测控制的能量管理实时优化策略，比基于传统多步神经网络策略的等效燃油消耗量减少了3.9%，改善了整车燃油经济性。

Energy Management Optimization for Hybrid Electric Unmanned Tracked Vehicles Based on Path Planning UsingCNN-LSTM Prediction

As one of the key technologies of hybrid electric vehicles, energy management is critical to the entire efficiency and fuel economy. As the driving cycle of unmanned tracked vehicles is uncertain, conventional energy management strategies must deal with new challenges. To improve the prediction accuracy, a prediction model based on Convolutional Neural Networks and Long Short-Term Memory (CNN-LSTM) is proposed for processing both planned and historical velocity series. An optimal forward dynamic programming algorithm is proposed to solve the optimal control problem of energy management. Based on the prediction results, a model predictive control algorithm is adopted to realize real-time optimization of energy management. The effectiveness of the method is proved by using collected data from actual field experiments of unmanned tracked vehicles. Compared with multi-step neural networks, the prediction model based on CNN-LSTM improves prediction accuracy by 3%. The energy management strategy based on model predictive control reduces fuel consumption by 3.9% compared to the traditional regular energy management strategy.