Why performance ratio differs between monitoring systems
Performance ratio: energy out ÷ (sunlight × size). Every source shows the same line. For the same plant and the same month, Source A reports 81.9 %, Source B 76.1 % and Source C 78.4 %. Nobody is wrong. Under each word sit choices nobody shows you.
The formula everyone agrees on.
PR = Energy out ÷ (Sunlight × Size)
what the plant delivered, divided by what the sun offered a plant this size
81.9 %
Source A
76.1 %
Source B
78.4 %
Source C
Three words, one division. Every source shows the same line. Under each word sit choices nobody shows you.
Synthetic example. The numbers are made up, the mechanics are not.
Which energy?
PR = Energy out ÷ (Sunlight × Size)
A, B and C are the three sources. Each badge marks what that source chose.
- Which meter? (the inverters sit before the transformer, the grid meter after it)
- sum of the inverters A grid meter B C
- How counted?
- counter difference C power × time A B
- A gap in the data? (ten minutes nobody received)
- counts as zero B filled from the day before A interval dropped C the counter bridges it
1.2
percentage points between the three sources, from the words “energy out” alone
+0.8
Source A
−0.4
Source B
±0.0
Source C
Shift in percentage points against one set of rules for all three. The grid meter sees the transformer losses, the inverters don't.
Synthetic example. The numbers are made up, the mechanics are not.
Which sunlight?
PR = Energy out ÷ (Sunlight × Size)
A, B and C are the three sources. Each badge marks what that source chose.
- Which source?
- sensor on site A C satellite B weather model
- Which sensor? (the plant has three)
- one, by name A average of all C average of those that pass the checks
- Which plane? (flat, or tilted like the modules)
- tilted, measured A C flat, converted B
3.6
percentage points between the three sources, from the word “sunlight” alone
+1.6
Source A
−2.0
Source B
−0.4
Source C
Shift in percentage points against one set of rules for all three. The biggest lever on the page, and the one nobody asks about.
Synthetic example. The numbers are made up, the mechanics are not.
One dirty morning.
Tuesday, 07:00 to 10:00: dew on the sensor, it reads 40 % low. Same reading, three sources, three honest days.
Source A
no check
used as is: weak sun, great PR
91 %
Source B
checked against satellite
satellite instead, for those three hours
80 %
Source C
checked: range and flat line
hours dropped, from both sides
82 %
Day PR, same Tuesday. None of the three is lying. Which rules made the number is hard to track.
Synthetic example. The numbers are made up, the mechanics are not.
Which size?
PR = Energy out ÷ (Sunlight × Size)
A, B and C are the three sources. Each badge marks what that source chose.
- Which number?
- module nameplate (kWp) A B as built, after the swap C inverter limit (kW)
- As of when? (four strings were replaced on the 12th)
- today's value, for the whole month B dated, as it was each day A C
- Ageing?
- none A B 0.5 % per year C
0.3
percentage points between the three sources, from the word “size” alone
±0.0
Source A
−0.3
Source B
±0.0
Source C
Shift in percentage points against one set of rules for all three. Small this month. Two percentage points in a repowering year, every month, until someone dates the master data.
Synthetic example. The numbers are made up, the mechanics are not.
Which hours?
The part nobody shows.
PR = Energy out ÷ (Sunlight × Size)
A, B and C are the three sources. Each badge marks what that source chose.
- Which hours count?
- all of them A B daylight ≥ 50 W/m² C
- Which month?
- calendar month, local time A C calendar month, UTC B
- One side missing? (energy without sunlight, or the other way round)
- keep what is there B drop both A C
- Temperature?
- not corrected A B corrected to 25 °C C
0.7
percentage points between the three sources, from the counted hours alone
+0.5
Source A
−0.2
Source B
−0.2
Source C
Shift in percentage points against one set of rules for all three. Each one small, none of them zero, all invisible in the formula.
Synthetic example. The numbers are made up, the mechanics are not.
Nobody is wrong. Nobody wrote it next to the number.
| Layer | Source A | Source B | Source C |
|---|---|---|---|
| one set of rules | 79.0 % | 79.0 % | 79.0 % |
| Energy out | +0.8 | −0.4 | ±0.0 |
| Sunlight | +1.6 | −2.0 | −0.4 |
| Size | ±0.0 | −0.3 | ±0.0 |
| Hours | +0.5 | −0.2 | −0.2 |
| reported | 81.9 % | 76.1 % | 78.4 % |
Almost six percentage points between the three, on a KPI that decides bonuses and sales. None of it stands next to the number.
Synthetic example. The numbers are made up, the mechanics are not.
Make the base row yours.
Audit
The KPIs your money depends on, recomputed by you from the raw readings. In this example, A's 81.9 comes from +0.8 for the inverter sum, +1.6 for one named sensor and +0.5 for dropping hours with one side missing.
Compare
One base row for every plant, whichever portal it runs on. Twelve assets, three providers, one number that means the same thing.
Let the portals monitor. They are good at it.
Questions
Why does the performance ratio differ between monitoring systems?
Because the formula hides choices: which meter, how gaps are handled, which irradiance source and sensor, which plane, which capacity as of which date, and which hours count. Each is defensible. None of them is written next to the number.
Which choice moves the number most?
In this example the word “sunlight”: 3.6 percentage points between the three sources, purely from what each means by it. It is the choice people ask about least.
Is one of the portals wrong?
No. Each one applies a defensible rule. The problem is that the rule is not written next to the number, so you cannot tell which rules made it.
How do I find out why my own portals differ?
Recompute the number from the raw readings with your own rules, written down once. When a portal differs, test which choices reproduce its number. Portals rarely publish their rules, so part of the gap may stay open. That is what the KPI audit does for one plant and one month.
Send one plant, one month.
You get performance ratio and availability as a base row with the rules written down, the difference against each reported number with what explains it and what stays open, a method card next to every number and a 30-minute walkthrough. Nothing changes in your monitoring.
Send one plant, one month