In recent years, rapid improvements in battery technologies in addition to decreasing prices have led to their widespread use in various applications. Due to their high energy and power density, efficiency and service life, lithium-ion batteries have become established in the vast majority of sectors. In addition to power tools and multimedia applications, this includes stationary applications of various sizes, as well as all levels from micro hybrid to fully electric vehicles. Depending on the application needs, lithium ion batteries therefore cover a wide range of energy and power densities as well as life spans.
Regardless of the application, it is crucial, to accurately determine various states of the battery while in use. This task is performed by means of a Battery Management System. Diagnostic algorithms on a BMS determine values like the State of Charge (SOC), State of Health (SOH) and State of Power (SOP) of the battery in order to ensure its safety and optimal utilization. Numerous algorithms for this purpose were developed in recent years. Typically, these are designed and validated for a specific application and certain environmental conditions. However, the vast majority of algorithms are affected by different power profiles, cell types, and BMS hardware, among other factors.
In order to investigate the aforementioned effects, a Model-in-the-loop toolchain for battery management systems has been developed and implemented. In addition to real, recorded driving data from various applications, simulated power profiles as well as synthetic current profiles can be used as input data. A high precision equivalent circuit battery system model based on both, electrochemical impedance spectroscopy and time domain measurements is used as the reference. Consequently, the diagnostic algorithms using more simplified battery models are tested against the reference. The toolchain includes the simulation of the full signal path through the BMS hardware and software. Thus, it is possible to test and compare diagnostic algorithms under the same, but freely definable boundary conditions.
In this work, the toolchain was optimized with respect to the input data, and the impact of different applications on the performance of the implemented state estimation algorithms was thoroughly investigated. For this purpose, state estimation algorithms such as improved Coulomb counting and Kalman filters were tested under different conditions. Input data were obtained from a vehicle database, a vehicle model and synthetic profiles. In addition to evaluating the algorithms, the obtained results serve to validate the functionality of the toolchain with respect to different algorithms and applications.