Data Driven Approach for Sensorless Temperature Estimation for Lithium-Ion Cells based on Pulse Resistance Measurements

Thematic block:


Other authors:



Knowing the temperature distribution within a battery pack is crucial, because of the impact of temperature on capacity loss, power degradation and safety issues. Temperature measurements are usually realized with external temperature sensors attached to a few crucial cells throughout the battery pack, leaving the majority of cells in bigger battery packs unattended. Existing methods in literature use electrochemical impedance spectroscopy (EIS) for estimating the temperature of a cell. By utilizing the relation of a cell’s temperature and certain criteria like real/imaginary part, phase and/or frequency in the impedance spectrum of the EIS, these methods can estimate the temperature for all cells in a battery pack without external temperature sensors. The EIS based approaches still have the disadvantage of adding additional hardware, cost and weight to the battery system for generating the excitation for the EIS measurement.
This work presents a sensorless method for determining the temperature of a cell in a battery pack by exploiting the relation of the cell’s overpotential and its temperature using the example of an 18650 NMC-811 cell. In contrast to the EIS based methods, the presented method does not rely on additional hardware for generating an excitation. Naturally occurring current changes in the battery load are utilized as excitation. The overpotential change caused by load current change is related to the cell’s materials and processes, which again are dependent on temperature. Based on this relation, a method for estimating a cell’s temperature using direct current resistance (RDC) calculated after a certain time (dt) after the current change, was developed.
Besides the temperature dependence, the RDC is also dependent on other pulse and cell parameters. Therefore, a set of reference pulses at 10°C, 20°C, 30°C and 40°C was recorded to investigate the influence of state-of-charge (SOC), pulse rise time and duration as well as the pulse current amplitude and direction on RDC. Analyzing the reference pulses showed that a dt in the 100 ms regime showed the greatest sensitivity to temperature and the least dependence on other factors. The method was finally validated with several different driving cycles applied to a module of six serial connected cells under an externally applied constant temperature gradient. The results of the validation experiment showed that the method is capable of determining the temperature of each cell in the demonstrator module with a root mean square error below 1 Kelvin.

Would you like to contact this author?
We are happy to forward your request / feedback.