"Range anxiety" remains one of the major obstacles to the wider adoption of electric vehicles (EVs). While range prediction technologies exist, most rely on simulated conditions or limited datasets, making it difficult to accurately capture variations caused by regional climate, road conditions, and vehicle types.
In a recent study published in Applied Energy, a research team led by Prof. CHEN Zhongwei and Associate Prof. MAO Zhiyu from the Dalian Institute of Chemical Physics (DICP) of the Chinese Academy of Sciences (CAS), in collaboration with Associate Prof. ZHANG Zhaosheng from the Beijing Institute of Technology, developed a comprehensive real-world data–driven framework for EVs range prediction and intelligent management. They have achieved a high-accuracy Remaining Driving Range (RDR) estimation under diverse, real-world operating conditions and provided an engineering-ready solution for intelligent EV fleet management.
To address the practical needs of real-vehicle applications, the researchers constructed an integrated framework for online range estimation and optimization analysis. This framework incorporates multiple sources of influence, including driving behavior, ambient temperature, and battery State of Health (SOH). Using a Random Forest algorithm, the researchers implemented a "two-step" estimation strategy—predicting energy consumption per unit distance, then deriving the remaining range. Compared with traditional "black-box" prediction methods, this step-wise modeling approach not only improves prediction accuracy but also enhances model interpretability, enabling researchers to quantitatively understand which factors influence the driving range and to what extent.
The remaining driving range estimation and analysis framework for electric vehicles (Image by ZHOU Litao)
Additionally, the researchers systematically validated the framework using three years of real-world operational data collected from passenger cars and buses operating in multiple cities, covering a combined driving distance of more than 300,000 kilometers. Results show that the average relative error between the predicted RDR and the actual drivable distance is below 5.5%, outperforming traditional methods.
Further analysis revealed that average current—reflecting the power intensity of the trip—and average speed are the key determinants of energy consumption. Optimization studies showed that reasonable adjustments in driving behavior could increase driving range by more than 30% for passenger cars and more than 10% for buses.
"Our study not only answers the question of 'how far the vehicle can go', but also provides a quantitative basis for 'how to go farther'," said Prof. CHEN. "It offers high-value technical support for EV fleet management, energy optimization, intelligent scheduling, and vehicle residual value assessment."