Twelve Solar Plants, Three Portals: Why the Numbers Don't Compare

Portfolios grow plant by plant, and each plant comes with the portal its builder or its inverter maker chose. After a few years and a couple of acquisitions, twelve plants report on three portals.

Then someone puts the twelve availabilities and performance ratios in one table and ranks the plants. The ranking compares the portals' rules as much as the plants.

Two plants, ranked twice

Synthetic example. The numbers are made up, the mechanics are not.

Portal A counts a data gap as available and weights every inverter the same, so North's logger trouble and the outage of its largest inverter weigh less. Portal B weights inverters by capacity and leaves gaps out. Computed the same way, South is ahead.

Nobody is wrong. Each portal applies a defensible rule. They are just not the same rule, and neither is written next to the number. Our page on comparable KPIs shows the same effect on one plant with three numbers.

What differs between portals

• The definition. For availability: which hours count, how inverters are weighted, what a data gap is, what is excluded and what "unavailable" means. For performance ratio: which energy, which sunlight, which size and which hours. See why performance ratio differs.
• The capacity. The DC nameplate from the module list, a field in the portal, the figure in the contract, sometimes the AC capacity. A plant entered at 9.6 instead of 10.0 MWp gets a performance ratio about 4 % higher, from the same energy.
• The time stamps. A value stamped 12:15 may cover 12:00 to 12:15 or 12:15 to 12:30. One portal stores local time, another UTC. On the day the clocks go back, the local day has 25 hours, and not every import keeps all of them. See Sunday had 25 hours.
• The units. Power or energy per interval; 5, 10 or 15 minutes; meter counters or interval values. A 15-minute power value in kW, read as energy in kWh, is four times too large.
• The names. The same sensor is "POA" in one export, "GTI" in another and "Sensor_1" in a third. And "irradiance" in a fourth may be measured on the flat, not in the plane of the modules.
• The adding up. A monthly performance ratio can be the energy of the month over the sunlight of the month, or the average of thirty daily ratios. Both are reasonable. They give different numbers.

Each of these is small on its own, and none of them shows in the final number. Together they can move a plant from the top of the table to the middle.

Why it gets harder as the portfolio grows

Every new portal brings its own format, names and rules. The people who knew a portal's quirks move on. The spreadsheet that maps portal A's names to portal B's is kept by one person. Each acquisition adds a portal, and each portal adds work to every comparison.

Doing it yourself

1. Write one definition per question: technical availability, contractual availability per contract, performance ratio. Once, for every plant.
2. Pick one source for the capacity and the plant data, and keep it with the dates it was valid. See When a Solar Plant Changes, Its KPIs Move.
3. Bring every portal's raw data into one format: interval start, UTC, energy per interval, one name per measurement, and a gap left as a gap.
4. Compute every plant from its raw data with the same rules, instead of taking any portal's KPI.
5. Keep a list of each portal's quirks. It will grow.

The format work is the larger part, and it starts again with every new portal.

Or let us do it

Let the portals monitor. They are good at it. We compute one base row for every plant, whichever portal it runs on: its KPIs from its raw readings, with the same written rules for every plant. You decide the rules, and a rule applies to every plant at once, so twelve plants on three portals still give one number that means the same thing. Start with one plant, one month.

See also

• Checking the Availability in an O&M Report: What It Takes
• Can You Get Your Own Solar Data Out?