How long will your battery really last?
Weltrom forecasts how long UPS and storage batteries will last, using the data their battery management system (BMS) already records. Move the sliders, or analyse your own data file. Nothing leaves your computer.
How to read the chart: the line is the expected health. Half of all similar batteries stay inside the shaded band and reach end of life within the thick stretch. How the forecast works
The window is wide for a typical battery because batteries of the same type age at quite different speeds. With your own data, the model measures your battery's speed and the window gets narrower. Each chemistry follows its own typical curve shape from published research. Most lithium batteries lose capacity fastest at first and then slow down, because a protective layer inside the cells grows ever more slowly (Attia et al. 2022). Lithium iron phosphate keeps slowing down (Naumann et al. 2018) and aged slowest of all in long-term tests below 80 % (Sandia, Wittman et al. 2023). NMC and especially LMO/NMC cells tend to speed up again late in life, more so when kept fully charged. Real batteries vary: in the Sandia tests some cells sped up and others never did. Today's capacity loss comes from 21 measured home-storage lithium batteries (Figgener et al. 2024), or from your own data. Temperature and charge-level effects use a published aging model for lithium iron phosphate cells (Naumann et al. 2018). Indicative only.
Analyse your own data
Upload a data export from your battery management system (BMS) or energy management system (EMS). You get the battery's measured state of health (SOH), capacity loss per year and end-of-life (EOL) estimate, with the method and uncertainty behind every number. The sliders above then switch to your battery.
Your data stays on your computer. The analysis runs inside this browser tab, and nothing is uploaded to a server. The first run downloads the analysis engine (about 10 MB); after that your browser keeps it.
These values now drive the forecast above. Move the sliders to try "what if" on your own battery.
Full technical report
Which file formats work?
An energy management system (EMS) "Professional parameters" style export (.xlsx); a CSV file with the same column names
(Time, Battery voltage(V), Battery current(A), SOC(%) = state of charge, min/max cell voltage and IDs…);
or a CSV in our template layout: timestamp, pack_v, pack_i_a, soc_pct plus optional
cell_v_min, cell_v_min_id, cell_v_max, cell_v_max_id, cell_t_min_c, cell_t_max_c, asset_id.
Several batteries in one file are reported separately (by asset_id).
About the sample data
Two years of a real home-storage lithium iron phosphate (LFP) battery: system 14 of the RWTH Aachen home-storage dataset (Figgener et al. 2024, Nature Energy, doi:10.1038/s41560-024-01620-9; data CC-BY-4.0, doi:10.5281/zenodo.12091223), reduced to 5-minute values. A field capacity test 3 days after the end of the sample measured 89.4 %.
What is this, exactly?
A preview of BatteryMind, Weltrom's battery-health analytics in development. Its methods are validated against 60 real field capacity tests. The validation data is lithium home and grid storage, not yet UPS fleets. Results are indicative and don't replace a capacity test. Methods, open data and validation →
From raw BMS signals to a replacement decision
No new sensors, no hardware swap. Weltrom sits on top of the telemetry your UPS already produces.
Ingest
Pull BMS/UPS telemetry via standard protocols (Modbus, SNMP, API) — independent of vendor or chemistry, VRLA or Li-ion.
Model
Apply physics-grounded aging models — not black-box ML — to turn raw signals into State of Health, recoverable capacity, and a failure-risk trajectory per string.
Alert & recommend
Get a ranked list of strings that need attention, weeks before a threshold would be crossed — each with a specific action: rebalance, monitor, or replace.
UPS batteries are managed with the least information of any critical asset in the building
They sit between your facility and a blackout — yet most teams have no reliable way to know their real condition until something forces the question. That gap shows up as two distinct, expensive failure modes.
Unexpected failure
A string looks fine on the BMS status light — right up until it's called on to hold load, and can't. BMS dashboards report pass/fail against fixed thresholds, not degradation trends, so early warning signs — rising internal resistance, float current drift, cell imbalance — go unseen until it's too late to act.
Blind replacement
With no reliable way to see actual remaining capacity, facilities fall back on fixed replacement intervals — typically 3–5 years for VRLA — regardless of real condition. Some strings are pulled with years of life left; others degrade faster than the schedule assumes and fail before their "expiry" date.
Both failure modes trace back to the same root cause: a BMS is built to protect the battery moment-to-moment, not to tell you how it's aging over months and years. That's a fundamentally different problem — and it's the one Weltrom solves.
Want this forecast for every battery you run?
Weltrom connects to your existing BMS data and keeps every string's health, aging rate and end of life up to date, with an alert before anything crosses a threshold.