Research on Calendar Aging for Lithium-Ion Batteries Used in Uninterruptible Power Supply System Based on Particle Filtering

Overview

The study asks how SOH and remaining useful life can be estimated accurately for UPS batteries, where slow, hard-to-measure calendar ageing dominates. Capacity fade is modelled as y_k = A(SOC)·exp(-Ea/RT)·k^z (k in days), with z = 0.5 if fade is driven only by SEI growth. Because SCADA showed constant ambient temperature and float voltage, the model reduces to y_k = α·k^z. A particle filter estimates SOH and RUL for each cell, and pack and group RUL are taken as the minimum over series-connected cells.

The UPS cells are 100 Ah, 3.6 V LFP; a battery group has seven packs and each pack has twelve series cells. The calendar-ageing protocol was CC-CV charge at 0.3 C (30 A) to 3.6 V per cell or 580 V per pack, storage at 25 °C with recharge if a cell drops below 3.35 V, and capacity checks every half month. Thirteen capacity-verification points were collected over about half a year.

Results: per-cell SOH estimation RMSE ranged from 0.15% to 9.07% (MAPE 0.16% to 6.85%). With end of life defined as 50% of nominal capacity, per-cell RUL estimates ranged from 194 to 1950 days, and the pack RUL (minimum cell) was 194 days. One weak cell can therefore set the replacement date for an entire string even when the other cells have years of life left.

Limits: only about half a year of test data and an assumption of constant temperature and float voltage; long-term predictions rest on a very short data window and the RUL range (194-1950 days) shows large uncertainty. The particle filter's adjustable parameters were tuned for UPS LFP batteries with a grid-search method.

Key points

1. Calendar ageing is modelled as y_k = A(SOC)·exp(-Ea/RT)·k^z (k in days); z = 0.5 if fade is driven only by SEI growth. At constant temperature and float voltage the model reduces to y_k = α·k^z.
2. The UPS cells are 100 Ah, 3.6 V LFP; a battery group has seven packs and each pack has twelve series cells.
3. Protocol: CC-CV charge at 0.3 C (30 A) to 3.6 V per cell or 580 V per pack, storage at 25 °C with recharge if a cell drops below 3.35 V, capacity checks every half month.
4. Thirteen capacity-verification points were collected over about half a year.
5. Per-cell SOH estimation RMSE ranged from 0.15% to 9.07% (MAPE 0.16% to 6.85%).
6. With end of life defined as 50% of nominal capacity, per-cell RUL estimates ranged from 194 to 1950 days and the pack RUL (minimum cell) was 194 days.

Using it in design

• Plan UPS battery replacement from the RUL of the weakest cell rather than the string average; in a series string one cell sets the replacement date for the whole string even when others have years of life left.
• Track float-charged LFP batteries' calendar ageing with a power-law model (day^0.5 type) and periodic capacity verification; add cell-level SOH and RUL estimation to the BMS.
• Report RUL predictions based on a short data window with their uncertainty range (for example 194-1950 days) and update them with new capacity measurements.

Who should read it

Facility engineers planning UPS battery maintenance and replacement, and BMS developers.

Document contents

1. 1. Introduction (1.1 Research Background and Motivation, 1.2 Literature Review, 1.3 Main Contributions)
2. 2. Battery Calendar Aging Modeling
3. 3. Methodology (3.1 Battery Cell SOH Estimation Algorithm, 3.2 Battery Pack and Group SOH Estimation Algorithm, 3.3 Battery Cell RUL Estimation Algorithm, 3.4 Battery Pack and Group RUL Estimation Algorithm)
4. 4. Results and Discussion (4.1 Experimental Testbench, 4.2 Results and Analysis: battery SOH estimation, battery RUL estimation)
5. 5. Conclusions

Related standards

• IEC 62040-3

Read the document

Publication details

Authors

Wei Xu; Hongzhi Tan

Publisher

World Electric Vehicle Journal (MDPI AG)

Year

2023

Volume

14

Issue

8

Pages

209

Document language

English

Source type

Scientific

Access

Open licence — downloadable

DOI

10.3390/wevj14080209

Licence

CC BY 4.0 — share with attribution

Licence evidence

www.mdpi.com

Citation