Beyond daily kilowatt-hours: five solar monitoring metrics that find real problems

Daily kilowatt-hours are the number every monitoring app puts on its home screen, and they’re the least useful number for judging whether a solar system is healthy. A low day might be a broken inverter or just a gray Tuesday in Lexington. Good solar performance ratio monitoring separates the two. Track five things instead: a weather-corrected performance ratio, availability, specific yield against an expected baseline, the degradation trend, and the fault signatures (clipping, inverter-level anomalies, soiling, snow and shading) that show up in interval data.

None of this needs special equipment. It needs a full data export, a reasonable baseline and the patience to compare like with like. Here’s how each metric works, a worked example from one illustrative Kentucky year, and what to hand a licensed technician when the numbers point to a real problem.

Why the daily total can’t tell you much

Production depends on two things: how much sunlight reached the panels and how well the system turned it into electricity. The daily total blends them. A system can lose 15% to a failed string and still post a “record day” in May, because May is sunny. And a perfectly healthy system will look sick in a cloudy December.

Every useful metric below does the same thing in a different way. It divides out the weather so what’s left is the system.

Solar performance ratio monitoring: correct for the weather

The performance ratio (PR) is defined in the international monitoring standard IEC 61724. It compares the AC energy the system delivered with what the panel nameplate rating would predict for the sunlight that actually hit the array. A national-lab report on the method describes PR as capturing losses from inverter inefficiency, wiring, mismatch, heat, soiling, downtime, shading and component failures. That’s most of what can go wrong, in one number.

The same report notes that PR values for new systems typically range from about 0.6 to 0.9. The catch is temperature. Panels lose output as they heat up (the report cites roughly 0.5% per degree Celsius), so PR runs high in winter and low in summer. The report notes seasonal swings as large as about ±10%, enough to let a poor system pass a winter test and a good one fail a summer test.

You can see this in a model of a healthy array. We ran the PVWatts V8 model for a 1 kW, south-facing, 25-degree array in Lexington with default losses. The modeled PR is about 0.84 in January and 0.73 in July, a drop of nearly 14% with nothing broken. For the full year it’s about 0.77.

Two practical rules follow:

• Compare like months. This July against last July, not July against March.
• Use a weather-corrected PR when you can. The national-lab method adjusts each period to the site’s typical average cell temperature, which removes most of the seasonal bias. Larger systems with on-site irradiance and temperature sensors can do this directly. Homeowners can approximate it with satellite irradiance data, with more uncertainty.

Availability: was the system actually running?

Availability asks a blunt question: what share of the energy the system could have produced was lost because equipment was off? A 2024 national-lab analysis of the PV Fleet Performance Data Initiative defines it in energy terms at the inverter level. It counts production lost while an inverter was completely offline. Partial reductions from clipping or heat don’t count against it.

In that study of 1,128 mostly commercial-scale systems, median steady-state availability was 0.991 and the 10th-percentile system was 0.947, after excluding a first six months in which availability ran lower. Larger systems tended to do slightly worse than smaller ones.

For a typical home system, the arithmetic is simple. If an 8 kW array in Lexington sits dead for nine days in June, it loses roughly 335 kWh. PVWatts budgets 3% a year for availability losses, about 325 kWh for that system, so nine days in June uses up the whole year’s allowance.

One caution: communication gaps are not outages. If your home Wi-Fi drops, the portal can show zeros while the system keeps producing. The same report treats a zero data channel as a gap, not a loss, when the cumulative energy meter shows production continued. Check the lifetime energy counter, or your utility’s record of exports on sunny afternoons, before you assume the worst.

Specific yield against a PVWatts baseline

Specific yield is annual production divided by system size, in kWh per kW of DC capacity. It lets you compare a 6 kW system with a 12 kW one, or your system with a model.

The free reference baseline is PVWatts, now published by the National Laboratory of the Rockies (formerly NREL). Version 8 uses typical meteorological year (TMY) data from the National Solar Radiation Database. For our Lexington test case (south-facing, 25-degree tilt, default 14% system losses), PVWatts estimates about 1,355 kWh per kW per year.

Three things to keep in mind:

• TMY is typical, not a forecast. Actual years vary, and PVWatts shows an expected range based on 30 years of historical weather for that reason. One year 5% below baseline is not a finding.
• Defaults are assumptions. PVWatts assumes 2% soiling, 3% availability loss and 0% snow unless you change them. A tree-lined roof in Frankfort needs a different shading input from an open field in Bourbon County.
• Compare with the proposal, too. If your installer’s estimate was well above a reasonable PVWatts run for the same roof, the “underperformance” may be an optimistic promise, not a broken system. Our post on solar proposal red flags covers how those estimates get inflated.

Degradation: a trend, not a snapshot

Panels lose a little output every year. A widely cited national-lab review of nearly 2,000 reported field degradation rates found a median degradation rate of 0.5% per year and a mean of 0.8%, with 78% of reported rates below 1% per year. More recent fleet data put the median system-level rate at about 0.75% per year.

The compounding is gentle at those rates and severe above them:

Annual rate	Output lost after 10 years	After 25 years
0.5%	4.9%	11.8%
0.75%	7.3%	17.2%
2.0%	18.3%	39.7%

You can’t measure degradation from one year of data. Weather, outages and soiling all move a single year’s number by more than the degradation does. Analysts look at year-over-year changes in weather-normalized production across several years, after filtering out outages. If someone says your panels are “degrading fast” after 14 months, ask for the method.

Signatures: clipping, inverter anomalies, soiling, snow and shading

Hourly or 15-minute data shows patterns that daily totals hide. These are the common ones.

Signature	What it looks like	Usually means	Who confirms it
Clipping	Flat top at midday on clear days, at the inverter’s rated output	Usually a deliberate sizing choice (PVWatts’ default DC-to-AC ratio is 1.2). A flat top below the usual level suggests heat derating or a setting	Installer or qualified technician, if the level changed
Lagging inverter or panel	One unit runs consistently below identical neighbors	Shading, a failing module or connection, or a fault	Qualified technician
Soiling	Slow decline over dry weeks, then a step back up after heavy rain	Dust, pollen or droppings, worse on low-tilt panels near fields or gravel roads	You, from the ground; cleaning by a professional if needed
Snow	Whole array near zero for a day or two after snowfall, then recovery	Normal in Kentucky winters and rarely worth acting on	No one, unless it persists
Shading	A dip at the same clock time on every clear day, deepening in late fall	Trees, chimneys or new construction, often growing year by year	You, with a photo at the time of the dip

A worked example: one Lexington year, month by month

Here is an illustrative 8 kW system with made-up production data to show the method. “Typical” is the PVWatts TMY estimate. “Sun vs. typical” is the actual month’s irradiance compared with TMY, the kind of figure you’d get from a satellite irradiance service. “Adjusted” scales the typical estimate by that factor.

Month	Typical (kWh)	Sun vs. typical	Adjusted (kWh)	Measured (kWh)	Measured ÷ adjusted
Jan	671	96%	644	515	80%
Feb	714	103%	735	713	97%
Mar	865	94%	813	789	97%
Apr	980	105%	1,029	998	97%
May	1,112	98%	1,090	1,046	96%
Jun	1,114	102%	1,136	772	68%
Jul	1,113	101%	1,124	1,069	95%
Aug	1,130	104%	1,175	1,058	90%
Sep	970	97%	941	913	97%
Oct	902	106%	956	918	96%
Nov	698	93%	649	571	88%
Dec	576	99%	570	490	86%
Year	10,845		10,862	9,852	91%

The annual total is about 91% of typical. Most owners would shrug and blame the weather. But the weather-adjusted expectation is almost exactly typical, so the weather didn’t cause the shortfall. The monthly pattern tells the real story:

• January is snow, which is normal and not worth a service call.
• June is an inverter that sat offline for about nine days. That pulls this year’s availability to roughly 96.7%, well below the fleet median.
• August drifts down through a dry spell and recovers after September rain, which is a soiling signature.
• November and December sag together, which fits new shading as the sun drops lower. A clear-day hourly chart would confirm it with a dip at the same time each afternoon.

Fix the June problem alone and the year moves from 91% to about 94% of adjusted expectation.

How to export your data from any monitoring portal

Menu names differ by manufacturer, but the steps are similar:

1. Find the export. Look for “reports,” “download” or “export.” Many portals offer CSV files. Some limit history or resolution on homeowner accounts, so ask the manufacturer what an owner account can access.
2. Take the finest resolution available, ideally 15-minute or hourly, for the system’s full life.
3. Get device-level data too: each inverter, microinverter or optimizer, not just the system total.
4. Save the event or alarm log with timestamps.
5. Pull your utility data. A net meter records imports and exports, not total production, but exports on sunny afternoons confirm the system was running.
6. Gather the paperwork: the proposal’s production estimate, the equipment list and the system layout.

If your installer closed and the account was in its name, the equipment manufacturer may be able to transfer it to you. Our guide to what to do when your solar installer went out of business walks through it.

What to hand a licensed technician

We analyze data. We don’t climb roofs, open equipment or perform repairs. On-site diagnosis belongs to a qualified solar technician or licensed electrician. A good data summary makes their visit shorter and cheaper:

• System specs: module and inverter models, counts and string layout
• Dates and length of each outage, with alarm codes
• Which inverter, string or panel lags, by how much and since when
• Time-of-day patterns for suspected shading or clipping changes
• What you’ve already ruled out (weather, communication gaps, snow)
• A specific request: “Diagnose inverter 2’s June outage and quote the repair,” not “check everything”

Ask whether the fix is covered by an equipment warranty before you approve paid work. Don’t open inverters, disconnects or combiner boxes yourself.

What we’d do

Pull the full history, build a weather-adjusted baseline, and score the five metrics month by month. That’s the core of our System Health Check, which also covers orphaned systems whose installer is gone. For owners with several sites, Portfolio Performance Reporting tracks the same metrics monthly or quarterly so a dead inverter shows up in weeks, not at year-end. Either way, you get a short list of what to ask a qualified technician, and what not to pay for.

As of November 1, 2026: the national lab that publishes PVWatts was renamed the National Laboratory of the Rockies on December 1, 2025. PVWatts Version 8 now runs at pvwatts.nlr.gov and uses 2020 TMY weather data. Older nrel.gov links in proposals and installer paperwork may redirect or stop working.