当前位置: X-MOL 学术Mech. Syst. Signal Process. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Prediction of vehicle driving conditions with incorporation of stochastic forecasting and machine learning and a case study in energy management of plug-in hybrid electric vehicles
Mechanical Systems and Signal Processing ( IF 7.9 ) Pub Date : 2021-03-05 , DOI: 10.1016/j.ymssp.2021.107765
Yonggang Liu , Jie Li , Jun Gao , Zhenzhen Lei , Yuanjian Zhang , Zheng Chen

Prediction of short-term future driving conditions can contribute to energy management of plug-in hybrid electric vehicles and subsequent improvement of their fuel economy. In this study, a fused short-term forecasting model for driving conditions is established by incorporating the stochastic forecasting and machine learning. The Markov chain is applied to calculate the transition probability of historical driving data, by which the stochastic prediction is conducted based on the Monte Carlo algorithm. Then, a neural network is employed to learn the current driving information and main knowledge after the simplified correlation of characteristic parameters, and meanwhile the genetic algorithm is adopted to optimize the initial weight and thresholds of networks. Finally, the short-term velocity prediction is achieved by combining them, and the overall performance is evaluated by four typical criteria. Simulation results indicate that the proposed fusion algorithm outperforms the single Markov model, the radial basis function neural network and the back propagation neural network with respect to the prediction precision and the difference distribution between expectation and prediction values. In addition, a case study is conducted by applying the built prediction algorithm in energy management of a plug-in hybrid electric vehicle, and simulation results highlight that the proposed algorithm can supply preferable velocity prediction, thereby facilitating improvement of the operating economy of the vehicle.



中文翻译:

结合随机预测和机器学习的车辆行驶状况预测以及插电式混合动力汽车能源管理的案例研究

对短期未来驾驶状况的预测可有助于插电式混合动力汽车的能源管理并随后改善其燃油经济性。在这项研究中,通过结合随机预测和机器学习,建立了驾驶条件的融合短期预测模型。应用马尔可夫链计算历史驾驶数据的转移概率,并基于蒙特卡洛算法进行随机预测。然后,利用神经网络对特征参数进行简化关联后,学习当前的驾驶信息和主要知识,同时采用遗传算法对网络的初始权重和阈值进行优化。最后,通过将它们组合在一起,可以实现短期速度预测,整体表现是通过四个典型标准进行评估的。仿真结果表明,所提融合算法在预测精度和期望值与预测值之间的差异分布方面均优于单一马尔可夫模型,径向基函数神经网络和反向传播神经网络。另外,通过将构建的预测算法应用于插电式混合动力汽车的能量管理中进行了案例研究,仿真结果表明,该算法可以提供较好的速度预测,从而有利于提高车辆的运行经济性。 。径向基函数神经网络和反向传播神经网络的预测精度以及期望值和预测值之间的差异分布。另外,通过将构建的预测算法应用于插电式混合动力汽车的能量管理中进行了案例研究,仿真结果表明,该算法可以提供较好的速度预测,从而有利于提高车辆的运行经济性。 。径向基函数神经网络和反向传播神经网络的预测精度以及期望值和预测值之间的差异分布。另外,通过将构建的预测算法应用于插电式混合动力汽车的能量管理中进行了案例研究,仿真结果表明,该算法可以提供较好的速度预测,从而有利于提高车辆的运行经济性。 。

更新日期:2021-03-05
down
wechat
bug