Estimating the energy consumption and driving range of electric vehicles with machine learning

The Data-driven methods have been widely used in the SOC, SOH and energy estimation of electric vehicles, and they are recognized as the most promising approaches. However, the popular machine learning methods used for electric vehicles are often “black boxes” which result in the poor interpretation of the model. In this paper, a highly efficient gradient boosting decision tree (LGBM) is proposed to accurately estimate the driving range of electric vehicle. In this model, the feature importance scores are provided to discover the relationship. In addition, the proposed LGBM-T model is capable of generalizing the abstractions by taking into account the key features of time and temperature. Experimental results illustrate that the proposed LGBM-T algorithm is able to reproduce the driving mileage trajectory, with a low mean absolute error (MAE) bounded by 1.681% on real-world vehicles under complex operating conditions. The availability of this algorithm is further corroborated by comparing to the support vector machines (SVM) based estimator.