BU-901b: How to Measure the Remaining Useful Life of a Battery

All products have a Remaining Useful Life (RUL), governed by State-of-Health (SoH). This also applies to batteries, and better SoH assessment will improve RUL estimations. Batteries seldom fail unexpectedly; most reach end-of-life following the SoH trail. Capacity is the leading health indicator that fades linearly and predictably. Anomalies do occur; many are mechanical in nature caused by aggressive use that may lead to increased internal resistance (Ri) or electrical short. Dendrites in Li-ion are an example.

Establishing RUL requires a Minimum Viable Performance (MVP) by establishing the lowest acceptable function level below which the battery is no longer viable for a given application. End-of-life for most batteries is a capacity of 80%. Instead of a 10-hour service, as possible with 100% capacity, 80% will only give 8 hours of run time.

The starter battery in a vehicle provides more tolerance and cranking is still possible at a capacity of 30%, or lower. Passing a capacity test reading above 40% promises one year grace, good to the next service. Figure 1 demonstrates the capacity drop of a starter battery with end-of-life point at 30%.

More accurate RUL estimations are possible by tracking the SoH of a battery with cloud analytics. This is made possible by storing performance data received from diagnostic chargers, analyzers, monitors and rapid-testers over time.

Industries benefiting from cloud analytics are health care, public safety, defense, logistics, drones and robotic operators by tracing the capacity loss over time and estimating the time dropping to the red line denoting end-of-life. The red line denoting MVP is set by the intrinsic Target Selector.

A battery should have 20% remaining charge at the end of a day. If consistently low, the Target Selector should be set higher to secure enough capacity for unexpected events. However, with ample SoC, the threshold can be lowered to keep batteries in service longer. Observing the residual charge resembles an airline pilot carrying enough fuel to enable a secure landing with headwind.

Methods of Battery Testing

Battery testing and diagnostic evaluations vary according to battery system and application. To estimate RUL, capacity readings must be tracked over time, and this is a challenge with larger systems. Most Battery Management Systems (BMS) measure voltage, Ri and temperature. These parameters alone are unable to provide a capacity reading; however, with historic data and known MVP, URL can be predicted over time. Most common battery test methods are:

1. Analyzing big data with Artificial Neural Networking (ANN). Date stamp, load patterns and environmental stresses are added to the ANN data. This method does not test the battery by electrochemical evidence but collates peripheral data of large battery storage system (BSS).

2. Assessing the integrity of a battery by electrochemical impedance spectroscopy (EIS) with a frequency scan and analyzing the Nyquist plot by artificial intelligence. EIS can assess SoH of individual battery modules with a matrix derived from batteries with diverse SoH.

3. Extraction of SoH data from a BMS. URL from a conventional BMS is based on cycle count, depth of discharge and coulomb counting if deep-cycling is involved. A BMS will find anomalies but RUL assessment is limited without knowing the usable battery capacity.

4. Reading the Full Charge Capacity (FCC) of a SMBus battery. FCC represents the digital capacity based on coulomb counting during charge and discharge in the field. The data is readily available but inaccuracies occur with random use that can be corrected with calibration. See BU-603: How to Calibrate a “Smart” Battery.

5. Applying a full cycle with a battery analyzer, a service is best suited for batteries used in portable applications. Modern battery analyzers calibrate smart batteries, prepare packs for service after storage, get them transport-ready with AirShip and check the “chemical capacity” before replacement.

Cloud Analytics

Modern chargers include diagnostic features that assess the capacity by reading the FCC of a SMBus battery. Regular batteries use the parser technology by establishing the residual state-of-charge (SoC) with intelligent filtering and then measuring the coulombs required to fully charge the pack. SoC plus charge equals the usable capacity. The parser needs a long “runway” to measure the capacity; a topping charge alone cannot give a reliable reading.

Diagnostic chargers are hybrid devices that service smart and dumb batteries side-by-side. A SoH light verifies that the capacity is met, adjustable by the intrinsic target selector. The SoH light only illuminates if diagnostic data is available. SoH analytics in modern diagnostic chargers run in the background.

Conclusion

Cloud connectivity paves the way to Reliability Assurance Maintenance Systems (RAMS) in making battery fleets transparent to users and fleet supervisors alike. Battery diagnostics has been lagging behind other services, such as RCM (Reliability-centered Maintenance) that predicts mechanical wear-and-tear. Batteries can no longer be installed and forgotten. Knowing the URL enables scheduled replacement before a failure occurs. Modern diagnostic systems are available to grant risk management in batteries.