How to Predict Battery Remaining Useful Life from the BMS Data You Already Have

Most battery, inverter and UPS manufacturers already collect more data than they use. Every battery management system (BMS) logs voltage, current, temperature and charge cycles, often for years. Yet a field failure still usually shows up the same way: as a customer complaint, a warranty claim or an emergency service visit.

The good news is that you rarely need new hardware to change this. The signals that describe how a battery is ageing are already in the telemetry. This article explains how remaining useful life (RUL) prediction works, what data you need, and how to start.

State of health vs remaining useful life

Two numbers matter, and they answer different questions.

• State of health (SOH) tells you how healthy a battery is today, usually as a percentage of its original capacity.
• Remaining useful life (RUL) tells you how long it has left before it crosses a threshold you care about, such as 80% of original capacity or the point where it can no longer support the load.

SOH is a snapshot. RUL is a forecast, and the forecast is what lets service and sales teams plan ahead instead of reacting.

What data you need

You can usually start with what the BMS already reports:

1. Voltage and current over time, including how voltage sags under load.
2. Temperature, at pack level and ideally per cell or module.
3. Charge and discharge cycles, including depth of discharge.
4. Cell-level readings where available, which reveal imbalance between cells.
5. Context data such as installation date, model, and service and warranty history.

Data does not have to be perfect. Gaps, irregular sampling and different firmware versions are normal, and cleaning them is part of the work.

How the prediction works

A practical RUL model combines three layers:

• Feature extraction. Raw telemetry becomes indicators of ageing: internal resistance growth, capacity fade, voltage sag under load, thermal stress and cell imbalance.
• Degradation modelling. The model learns how each battery's indicators change over time and projects that trajectory forward, with a confidence band rather than a single date.
• Anomaly detection. Separately, the system flags batteries behaving unusually today (sudden temperature rise, abnormal current draw), which often explains early failures.

The output is not just a number per battery. It is a ranked list: which units to replace now, which to watch, and which are healthy.

From prediction to action

A forecast only creates value when it changes what teams do.

• Service teams move from calendar-based to condition-based maintenance, and visit sites before a failure rather than after it.
• Warranty and quality teams see which models, batches or operating conditions age faster.
• Sales teams get a renewal pipeline. A battery approaching end of life is a replacement conversation with a known date. Usage patterns also show which customers would benefit from an upgrade, such as more capacity or a solar add-on.

That last point is often overlooked. The same telemetry that predicts failures also tells you who is ready to buy.

How to start: a four-step pilot

1. Pick one product line with at least a year of telemetry and known failure or warranty outcomes.
2. Agree success criteria up front, for example how far in advance at-risk units are identified compared with what actually happened.
3. Back-test on history. Train on older data and check predictions against batteries that later failed.
4. Run live for a quarter with service and sales using the ranked lists, then decide whether to scale.

How Claysol helps

BatteryIQ is Claysol's battery health prediction software. It reads the BMS telemetry you already collect, forecasts state of health and remaining useful life, and gives service and sales teams ranked, actionable lists. It is licensed as software and works without new hardware.

If you make or operate batteries at scale, we can scope a pilot on your own data. Talk to our team.