Stochastic profiles from real driving data for offline battery state estimation validation
提出用马尔可夫链从真实驾驶数据生成随机功率谱的方法，分类城市、乡村和高速路况，用于验证三种电池荷电状态估计算法，发现不同路况下算法精度差异显著。
• Generation of driving cycles based on real-driving data by Markov chains • Driving cycle assessment by speed acceleration frequency distribution approach • Generation of modular power profiles for different driving conditions • Exemplary validation for state of charge estimation for selected driving cycles A crucial aspect of electric vehicle integration into sustainable transportation systems is the reliable estimation of battery states, such as the State of Charge, under realistic driving conditions to ensure the robustness of battery management systems. This work presents a method for generating stochastic power profiles using Markov chains based on real-world driving data of an electric vehicle. This data is classified into urban, rural, and highway segments, defining transition probability matrices for driving events like acceleration, deceleration, cruise, and idle. The resulting driving cycles serve as input for a vehicle model to simulate vehicle-specific power profiles, which are then used to validate three different state-of-charge estimators in a Model-in-the-Loop environment. Our results reveal significant differences in algorithm accuracy under various driving conditions. This work demonstrates the necessity for comprehensive validation using diverse, realistic profiles to ensure reliable battery state estimation, ultimately enhancing the understanding of algorithm performance.