Six kWh a Day Vanished

Read the two numbers together - that is the whole diagnosis.

• Voltages match (342 V vs. 345 V, 0.9% apart). A bypassed substring or a shorted module would drop String A's voltage by roughly a module's worth of Vmp (about 13 V per substring) and the MPPT would settle lower. It did not. So this is not an open substring, a bypassed module, or a missing module.
• Currents do not match (6.8 A vs. 9.4 A, a 28% deficit on String A). With voltages equal and irradiance equal, a current-limited string points to a series-resistance problem inside String A.
• Rule out the easy explanations first: uniform soiling (the customer's description matches String B, which is fine), no new shade, and no wiring change. If it were a string-level wiring or connector problem you would normally see a voltage anomaly or an intermittent arc-fault event as well.
• The thermal image closes it. 71 °C against 44–47 °C neighbours is a hot spot: a cell, solder joint, or junction-box connection dissipating power instead of passing it. That is consistent with a current deficit at normal voltage.

Conclusion: module-level failure in String A - cracked cell or failing interconnect/bypass diode - not a string wiring fault and not weather.

Quantify in watts, then in kWh, then in dollars.

• Actual array DC: (6.8 × 342) + (9.4 × 345) = 2,326 + 3,243 = 5,569 W.
• If String A matched String B: 9.4 × (342 + 345) = 9.4 × 687 = 6,458 W.
• Deficit = ~890 W at 980 W/m²; over roughly 6 peak-sun-equivalent hours that is about 5.3 kWh/day, which lines up with the observed 34–36 → 26 kWh/day drop (other days/conditions account for the remainder).
• At $0.18/kWh that is roughly $1/day in summer - and the hot spot is degrading, so the trend matters more than the current number.

Warranty file - assemble it properly:

1. IEC 62446-1 style commissioning/retest data: irradiance, module and ambient temperature, Voc, Isc/operating current per string.
2. Thermal images of the affected module with a temperature reference and an ambient reading.
3. A photo of the module serial/label and the array layout showing its position in the string.
4. A short written statement of the measured shortfall against nameplate at the recorded irradiance.

Do not simply swap the module and leave. Find the root cause - mechanical damage during installation, a shadow cast by a rail, a failing interconnect batch, or a connector - because the replacement will fail the same way if the cause is still there. Confirm the replacement module's Voc and current bin match the rest of the string.

Safety first, then the commercial message - in that order.

Interim safety:

• A hot spot is a localized power dissipator. It is not automatically a fire, but a progressing cell or solder-joint failure can arc or scorch a backsheet, and this one is already running ~25 °C above its neighbours.
• Where the inverter supports per-input isolation, take String A out of service until the module is replaced. If it does not, schedule the replacement promptly rather than "at the next service visit" and record the decision.
• Do not open junction boxes or probe module internals on the roof - the string is energized whenever the array sees light.
• Re-check the thermal image after the repair, and inspect the modules adjacent to it.

Customer message:

• State the measured numbers, not adjectives: "String A is producing 28% less current at the same voltage, and one module is running 25 °C hotter than its neighbours."
• State the impact in their terms: roughly 5–8 kWh per day, about a quarter of summer production, and worsening.
• State the plan and the owner of each step: warranty claim filed by you, module supplied under the manufacturer's product/performance warranty, labour covered per the warranty terms, and a target date.
• Do not promise a warranty outcome before the manufacturer responds - promise the documentation and the follow-up date instead.