With their high energy and high power density, lithium-ion batteries (LIBs) are widely used as power sources for electric vehicles and energy storage systems, and their demand continues to grow. However, the increasing energy and power density of LIBs leads to significant heat generation issues, and reflecting these safety concerns there have been a number of recent fires of high-performance batteries. Under these circumstances, it is increasingly important to have a reliable assessment of battery safety through precise diagnostics of battery state of health (SoH). As a matter of fact, the battery management system diagnoses and evaluates battery SoH by analyzing various parameters such as voltage, current, and surface temperature of battery, and among these there is no doubt that battery temperature is a critical safety factor. As LIBs become more high-performance and large-scale, the temperature distribution inside the battery becomes more uneven, which possibly degrades performance and eventually leads to failure, so it is very important to secure the uniformity of the battery temperature and at the same time to analyze the temperature inhomogeneity and sudden temperature changes inside the battery.A typical way to estimate battery temperature is to insert a temperature sensor into the battery or attach it to the battery surface. While these methods enable direct temperature measurement, inserting a sensor can mechanically affect the internal structure of the battery, and attaching a sensor to a surface can result in delayed detection of internal cell temperature changes and may not fully reflect the battery temperature especially if there is a large temperature distribution. As an alternative, temperature estimation using electrochemical impedance spectroscopy (EIS) has received significant attention. Although EIS analysis is an indirect method of obtaining internal cell information through the electrochemical signal, it has the great advantage that it is a non-destructive method that does not affect the cell structure, and the internal cell situation is fully reflected in the electrochemical signal. Nevertheless, typical temperature estimation methods using EIS are time-consuming because they require deriving temperature-related impedance components from impedance data measured over a wide frequency range (kHz to Hz). To solve this problem, Srinivasan pioneered a method that utilizes impedance values at a single frequency for temperature estimation, which uses a specific response signal that has a clear temperature dependence but is largely unaffected by other factors (such as state of charge (SoC)). While this method can effectively reduce the cost and time spent on temperature estimation using EIS and is one of the most promising methods of estimating cell internal temperature known to date, it has been reported that the accuracy of temperature estimation is low at high temperatures for reasons that are not yet fully understood.In this study, we diagnose the problems with the existing method of estimating the internal temperature of batteries through EIS and propose a way to improve the accuracy of temperature estimation at high temperatures. For this purpose, first, we measured the impedance at various SoCs and derived the characteristic frequency representing the reaction at the solid electrolyte interphase (SEI), which is the main reaction used for temperature estimation in the previous work. In addition, we thoroughly analyzed the impedance change with temperature to more clearly understand the resistance components contributing to the impedance at the characteristic frequency. The analysis showed that impedance value at characteristic frequency, which was thought to represent only one reaction at the SEI, contained signals from other reactions (e.g. interfacial charge transfer) at higher temperatures. This situation becomes more severe as the temperature increases, and is believed to be the reason why traditional impedance-based temperature estimation methods have limited accuracy at high temperatures. We also observed that increasing the characteristic frequency to avoid this signal overlap prevented signal disturbance due to charge transfer reactions, but instead introduced another signal disturbance due to stray inductance. Based on these analyses, this study proposes a new frequency value to be used for battery temperature estimation through EIS, and at the same time presents an impedance correction method that can reliably take only the impedance signal of a specific reaction with temperature dependence without interference from other reaction signals at that frequency. It is confirmed that the temperature estimation accuracy is greatly improved when using the method presented in this study compared to the existing temperature estimation method. In this presentation, we will critically discuss why existing methods have poor temperature estimation accuracy at high temperatures, and propose an impedance-based temperature estimation method that can achieve high accuracy over a wide operating temperature range.