How fast home batteries fade in real homes

The paper is "Multi-year field measurements of home storage systems and their use in capacity estimation" by Jan Figgener, Dirk Uwe Sauer and colleagues, published open access in Nature Energy on 16 September 2024. It is unusual in two ways. The batteries were privately owned systems doing their normal job, paired with rooftop solar, from 2015 to 2022. And the team didn't just trust the batteries' own readouts. They went out and ran 60 manual full-discharge capacity tests in the field, which they estimate took around 120 working days.

The headline number

In the authors' words, the systems "lose about two to three percentage points of usable capacity per year on average." The manual tests gave average fade of 2.1 points per year for one group of small systems (an LMO/NMC blend), 3.2 for medium NMC systems and 2.2 for medium LFP systems. The team's algorithm, which estimated capacity continuously from the operating data, gave slightly different averages for the same groups (3.1, 1.9 and 2.0 points per year). Two methods, same ballpark.

The spread matters as much as the average. The paper reports a range of about 2 points per year even among similar systems. The first systems hit the usual end-of-life line, 80% of nominal capacity, after about seven years, and some individual tests came in slightly below 80% after just 4.5 years. Others still looked healthy after the same time. If you extrapolate the slopes in a straight line, end of life lands anywhere from about 5 years (at 4 points a year) to 20 years (at 1 point a year), though the authors caution that a straight line is not reasonable for the slow agers, because fade tends to turn nonlinear toward end of life.

The part the cycle ratings miss

These batteries didn't work very hard. The medium NMC and LFP systems ran about 200 equivalent full cycles a year, and the small systems about 250. Most individual charge swings were shallow, under 20% depth of discharge, with a second cluster of near-full cycles. The pattern was seasonal: regular full cycles in spring and fall, fewer in summer (some systems were rarely emptied) and barely any in winter (some never filled).

Now do the arithmetic against a spec sheet. At 200 cycles a year, a cell rated for several thousand cycles would need decades to use them up. Yet these systems lost 2 to 3 points a year anyway. The paper normalizes fade per 100 full cycles too (1.2, 0.9 and 0.7 points for the three groups) but doesn't split how much of the loss came from cycling versus simply aging with time and temperature. My reading, and it is a reading, not the paper's claim: for a home battery, the calendar is at least as important a clock as the cycle counter. (Report 174 covers what a big cycle-count claim does and doesn't promise, and Report 114 covers heat.)

One more useful detail: usable energy fell about as fast as usable capacity. The authors conclude that losing capacity was the dominant aging effect, with any rise in internal resistance secondary over this period.

Why this didn't become a warranty war

Most systems still met their warranties. The reason is a quiet engineering choice: manufacturers state less capacity on the datasheet than the battery actually holds, building in an aging reserve. A battery can lose real capacity for years before it drops below the number printed on the label. The authors call it a positive sign that these first-generation products mostly held up. It also means the percentage you see in the first years can hide fade that is already happening.

Why your app's health number deserves skepticism

The paper's starting complaint is that owners have to rely on the battery management system's state-of-health figure, which has "unknown calculation schemes" and known inaccuracies. That is why the team built its own estimator and checked it against physical tests. Their method waits for the battery to reach full and empty naturally, lets the voltage relax, and counts the charge in between, correcting for the small current the battery's own electronics draw.

It worked best on chemistries with a sloped voltage curve. For LFP, now a common home-battery chemistry, the flat voltage curve made full and empty harder to pin down, and the estimates were the least precise of the three groups. Even the manual tests read LFP slightly low, because those systems didn't taper their current near empty during a full-power test. The measuring is hard, and how you test changes the number. (Report 150 covers why state of charge is an estimate, not a measurement.)

What this can't tell you

These were first-generation systems. They were measured from 2015 onward, and cells and management software have changed since. Treat the numbers as a baseline, not a prediction for a battery bought this year.

Twenty-one systems is a small sample, from three product types, all in German homes. Hotter climates and outdoor installs could age differently.

The authors have a commercial interest. Several are shareholders or employees of ACCURE Battery Intelligence, an RWTH spin-off that sells battery diagnostics, and the paper declares it. That doesn't undercut the measurements, which are published openly (about 146 gigabytes, 14 billion data points, on Zenodo), but it is worth knowing.

I care about this because I help design the AI battery-cycling systems for a veteran-owned (HUBZone) energy-storage integrator, run by fellow veterans, and honest health estimates are the foundation that kind of logic stands on. I do not own that company and earn nothing from this link. Full policy here. Nothing in this report involves them or any product they sell.