Comparative Analysis of Electric Vehicle Simulator for Accurate Battery Pack Internal Signal Generation

Abstract

The definition of accurate electric vehicle (EV) simulators can help mitigate the lack of large-scale public battery pack datasets in literature. This work compares two developed Simulink-based EV simulators that generate realistic EV battery pack signals from input driving sessions. The two EV simulators, referred to as simplified and advanced respectively, share the same architecture. However, they are equipped with internal blocks characterized by different complexity and precision. Both simulators generate time series of the vehicle's speed, and battery pack's current, state of charge (SOC), voltage, and internal temperature. Additionally, the simulators incorporate thermal and aging models, allowing for the emulation of a wide range of environmental conditions and aging statuses of the battery pack. A subset of inner parameters has been set, sourcing from online technical data sheets, to enable both virtual-EVs to mimic the same 2017 Volkswagen eGolf EV model. Indeed, given the availability of an acquired and ample real dataset specific to the same EV model, it is possible to perform an extensive and thorough validation of the simulated data. Both virtual-EVs prove to be accurate at simulating a battery pack under different aging conditions, although the comparison highlights the benefits of more sophisticated design choices, demonstrating the higher accuracy of the advanced virtual-EV over the simplified one. Indeed, the advanced virtual-EV achieves overall RMSE and R $^{2}$ values, for current, voltage, and SOC of 43.34A, 4.07V, 4.84% and 0.28, 0.93, and 0.96, respectively. The main design differences between the two virtual-EVs are presented, and, upon examining their computational burden, distinct utilization scenarios are proposed based on the user's needs.