A hybrid approach to estimating electric vehicle energy consumption for ecodriving applications

Estimation of the energy consumption rate for electric vehicles (EVs) plays a critical role in a variety of EV eco-driving applications and systems. As a result, quite a few studies have been dedicated to the development of estimation models. To improve the predictive and real-time performance of EV energy consumption rate estimation for eco-driving applications, we proposed herein a hybrid modeling approach which features knowledge-driven variable selection and data-driven statistical synthesis. Real-world driving data from a 2013 NISSAN LEAF were used to calibrate the models and evaluate their effectiveness in comparison to some existing models. Results from statistical analyses indicate that the proposed hybrid models outperform the knowledge-driven model or the data-driven model in terms of forecast accuracy.