Electric Vehicle Dynamic Operation Simulation with Data-driven and Physical-based Models

Aiming at the complexity of the dynamic operation simulation of electric vehicles (EVs), this paper proposes a dynamic operation simulation model that integrates data-driven and physical-based principles. This model framework combines the advantage of interpretability from the physical model while leveraging the strength of rapid simulation under dynamic operating conditions of the electric vehicles from the data-driven model. The physical model part covers key aspects such as vehicle dynamics modeling, regenerative braking system, temperature model, and battery state estimation model. The data-driven part extracts key features and labels based on actual vehicle operation data, and establishes a capacity-life prediction model for the power pack of an electric vehicle by using the long-short-term memory model (LSTM). By combining the physical model with a data-driven approach, this model effectively simulates dynamic changes in vehicle cabin temperature, battery pack temperature, and battery capacity degradation across varying operating conditions during their use lifetime in a short timeframe. Simulation and validation results based on real-world driving data show that the fusion model is highly accurate and reliable in both energy consumption prediction and battery life prediction, and can provide effective tools and theoretical support for performance analysis, optimization design, and energy management of electric vehicles.