High-precision energy consumption prediction is an important prerequisite for accurately predicting the running range of vehicle. A combined energy consumption model is established for a heavy-duty purely electric-powered vehicle operating on unstructured roads with significant load changes. The proposed model consists of two parts:a basic model for energy consumption calculation and a long short-term memory ( LSTM) neural network for difference correction. Based on the efficiency modeling of drive motor and transmission,the basic model is established in combination with vehicle driving dynamics. Then the LSTM neural network is used to correct the difference between the energy consumption prediction result of the basic model and the power test value of vehicle under typical operating conditions,which effectively improves the prediction accuracy of vehicle energy consumption under significantly variable loads and low signal-to-noise ratio gradient environments. Therefore,the combined energy consumption model has the advantages of simple parameters and model fitting without explaining the energy consumption laws. The real vehicle tests are analyzed. Compared with the VT-Micro model and the Radial basis function (RBF) neural network model for energy consumption,the average error rate of power prediction of the proposed combined model is reduced by 17. 76％ and 3. 35％,respectively,enabling the accurate real-time prediction of energy consumption for the heavy-duty purely electric powered vehicle under complex operating conditions.
维普
万方数据