Predicting the battery core temperature: Explanatory power of measurement quantities under different uncertainty scenarios

Abstract

Abstract Predicting the highest battery temperature, the core temperature, is an important task for the safe operation of lithium-ion batteries. This prediction task is complicated by inherent system uncertainties that result in uncertain core temperature estimates. Aside from model, parameter and measurement uncertainty, this also includes uncertain user behavior in form of uncertain future discharge currents. However, measurable quantities like voltage, surface temperature or discharge current can potentially decrease the uncertainty in predicting the core temperature. The extent to which a measurement is able to decrease this estimation uncertainty, called data worth, depends on the uncertainty scenario. We conduct a model-based study to investigate the potential of voltage, current and surface temperature measurements to decrease core temperature estimation uncertainty. We use our previously developed stochastic, physically-based battery model to estimate the core battery temperature of a cylindrical LiFePO4-Graphite cell. The data worth is computed with the Preposterior Data Impact Accessor method. We find that the common input to state-of-charge estimation methods, i.e. voltage and current measurements, can theoretically partially substitute a temperature measurement, if the user behavior is anticipated to some degree. Moreover, we highlight the importance of adequately estimating the involved uncertainties when assessing the data worth of measurement quantities.