Due to ongoing global warming and increasing air pollution, it is imperative to shift from fossil fuels to renewable energy sources. In this context, Electrical Energy Storage Systems play a central role in both grid stabilization and the long-term establishment of battery electric vehicles (BEVs) and consequently make a significant contribution to widespread decarbonization. Lithium-ion batteries are particularly suitable as energy storage devices for use in BEVs as they combine a high gravimetric energy density with high service life. To ensure the safety, reliability, and functionality of the battery system, battery management systems are used, and various internal states must be estimated.
Users of BEVs expect a long range according to the energy content of the battery as well as reliability and high user-friendliness. However, since the range of BEVs is typically less than that of vehicles with internal combustion engines, and charging stations are (as of today) less common than gas stations, an accurate prediction of the remaining driving range is essential for the acceptance of the vehicles. Physically, the remaining range correlates with the remaining usable energy, which depends on the future load profile as well as environmental factors until complete discharge. A determination of the remaining energy without the consideration of the aforementioned factors has several drawbacks, which lead to inaccuracies in range prediction: First, the energy losses due to the internal resistance are neglected. Second, the non-linearity of the voltage curve is not taken into account even though the extractable energy is not constant over a fixed-size SoC interval. Consequently, a so-called State of Energy (SoE) metric is introduced that inherently incorporates the voltage curve and environmental conditions.
In this work, different algorithms for the SoE estimation are presented and discussed. On one hand, the SoE can be determined according to its mathematical dependency on the SoC when only considering the non-linearity of the voltage curve. However, a disadvantage is that the performance of the prediction strongly depends on the estimation performance of the SoC. On the other hand, model-based algorithms have been implemented and tested in a Model-in-the-Loop Toolchain that additionally take into account energy losses due to cycling. Here, the estimation of the open-circuit voltage based on a battery model plays a vital role. For the validation of the implemented estimators, synthetic load profiles, as well as real-world driving profiles of BEVs, have been investigated. In a direct comparison of the estimators, the Unscented Kalman Filter in the scope of model-based estimators has proven to be a robust method for estimating the SoE.