基于LSTM神经网络的乘用车能耗预测

为改善当今石油供需矛盾和环境问题，针对乘用车提出了基于LSTM神经网络的燃油乘用车能耗预测模型。通过纵向动力学建模并匹配相应车型进行求解，结合GB/T 38146.1行驶工况数据，得出能耗随时间的变化率。构建LSTM神经网络架构，根据处理后的数据样本，对LSTM神经网络进行训练和评价。最后，通过LSTM神经网络和BP神经网络的仿真对比表明，随着迭代周期的增加，LSTM神经网络模型具有更高的精度，对能耗预测的准确性较好，对改善无人驾驶车辆的节能减排具有工程应用价值。

ENERGY CONSUMPTION PREDICTION FOR PASSENGER CAR BASED ON LSTM NEURAL NETWORK

In order to improve the contradiction of oil supply-demand and environmental problems, a fuel consumption prediction model for passenger cars based on LSTM neural network was proposed. Combined with the driving condition data of GB/T 38146.1, through the longitudinal dynamic modeling and matching with the corresponding vehicle model, the change rate of energy consumption with time was obtained. The LSTM neural network architecture was constructed, and the LSTM neural network was trained and evaluated according to the processed data samples. Finally, through the simulation comparison of LSTM neural network and BP neural network, with the increase of iteration period, LSTM neural network model had higher accuracy and better accuracy for energy consumption prediction, it was of engineering application value to improve the energy saving and emission reduction of unmanned vehicles.