Session: 07-01 Photovoltaic & Electrochemical Technologies

Paper Number: 112419

112419 - Predicting the State of Charge of a Battery at Rest Using the Open Circuit Voltage 

Lithium-ion batteries remain an important research topic due to their versatility in application, high power density, and potential role in reducing greenhouse gas emissions. As lithium-ion batteries are increasingly deployed, the breadth of use conditions and operating requirements continue to grow. This has led to the need to understand how various parameters affect the battery’s ability to ensure safe and reliable operation. One of the most basic and vital parameters describing the battery is the state of charge (SOC). Accurate estimation of SOC during use and during rest can be assessed by observing the voltage. For a battery at rest, the SOC is related to its open-circuit voltage (OCV). However, the OCV-SOC relationship is complicated and depends on the operation conditions before the battery is at rest. In addition, the voltage of a battery at rest varies for a period of time subsequent to the start of relaxation due to the diffusion of ions within regions with a concentration gradient. The voltage value taken after some duration of rest may not yet reflect the final voltage value that will be achieved at a state of equilibrium, but it is often referred to as OCV regardless. Therefore, the duration of relaxation of the battery also effects the ‘OCV’ – SOC relationship.

This work develops a methodology and simplified model for real-time applications that seeks to characterize the OCV – SOC relationship based on certain input parameters that capture the hysteresis effect of the OCV – SOC curve.  Extensive testing with lithium iron phosphate (LFP) cells shows that the charge verses discharge operation of the battery prior to rest is the most significant factor that impacts this curve. The C-rate (i.e., the rated current divided by the nominal battery capacity) and step change in SOC of the battery before relaxation also impact the curve to a smaller extent. Our model is developed by characterizing the battery through six different test procedures varying C-rate, step change of SOC, duration of rest and charge verses discharge operation of the battery. These tests allow for the complete characterization of the OCV – SOC curve. The model uses this information to predict the SOC of the battery. The OCV – SOC curve is divided into three sections, approximately 0-30, 30-60 and 60-100 percent SOC, with a different set of equations for each section. With the input of relaxed voltage, duration of relaxation and charge/discharge operation prior to rest, the model can predict the SOC within 2-3 percent. Compared to existing models with estimation errors in the range of 5 percent, this model offers improved accuracy, particularly for LFPs, which are more challenging to model. The SOC of the first and third section can be predicted more accurately due to the larger gradient in the curve, as compared to the second section where a few millivolts difference in OCV results in large differences in SOC. Providing additional data of C-rate prior to rest improves the accuracy of SOC prediction. Collected data shows that large differences in C-rate, such as 1C verses 0.1C, could affect the predicted SOC by 5 percent. Hence, this work provides a method to characterize the OCV – SOC relationship of a battery using a short test procedure and provides a model that can interpret this information to accurately predict the SOC of the battery based on the given OCV.

Presenting Author: Sahana Upadhya University of Wisconsin-Madison

Presenting Author Biography: Sahana Upadhya is a PhD student at the University of Wisconsin - Madison working in the Energy System Optimization group in the Solar Energy Lab. She is currently working on the characterization of batteries through state of charge and state of health estimation, with the aim of developing battery degradation models that can be integrated with hybrid renewable optimization models.