Data-Driven State of Health Estimation in Sparsely Sampled Lead-Acid Battery Packs
Open Access
- Author:
- Worman, Gabriel
- Graduate Program:
- Mechanical Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- March 20, 2023
- Committee Members:
- Robert Kunz, Professor in Charge/Director of Graduate Studies
Chris Rahn, Thesis Advisor/Co-Advisor
Jared Butler, Committee Member - Keywords:
- state of health
batteries
lead-acid
capacity
internal resistance - Abstract:
- Batteries degrade over time, so estimating their state of health (SoH) is critical to proper functioning of battery packs. Electric vehicles (EVs) in particular, can stall if the battery pack does not provide sufficient voltage to operate the electric motors (Vmin). The battery voltage depends on its state of charge (SoC), capacity (Q), internal resistance (IR), and discharge current (I). SoC and I are dynamic functions of time that vary during driving. Q and IR slowly change over time due to cell temperature and aging. In this work, three parameters associated with battery SoH are estimated from EV lab and field data, Q, IR, and cyclic Vmin. The data are collected from golf cars powered by six 8 V valve regulated lead-acid absorbent glass mat batteries. One EV is tested in a lab environment and the others are in use at a golf course. Data are sampled every five minutes for eight months and include max, min, average and instantaneous voltage and current, integrated current, and temperature. Q is estimated using discharge SoC and current throughput. IR is estimated using voltage and current during charging and compensated for temperature variations. Finally, the cyclic minimum voltage is tracked. The field and lab test data show aging trends for each of the three SoH parameters. Linear approximations are applied to these trends, with low mean absolute errors (MAE) seen for all batteries. Average MAE for Q is 11.83 Ah, with a standard deviation of 2.56 Ah. Average MAE for Vmin is 0.14 V, with a standard deviation of 0.064 V. Average MAE for IR is 3.31 m$\Omega$ with a standard deviation of 0.73 m$\Omega$. The lab data show similar trends, but with much smaller MAE.