Which String To Isolate First When Array Underproduces Sequencing Decision Tree

Why this matters

When an array underproduces, the temptation is to start pulling strings at random. On a multi-string inverter that wastes a sunny afternoon and risks live-DC errors. The right sequence is data-led: read what the monitoring and the inverter already tell you, then isolate the one string the evidence points at, confirm it, and only then open the next. A disciplined sequence finds a single failed string in one or two measured steps instead of a shotgun teardown, and it keeps you from de-energizing a healthy MPPT input that is carrying the load.

This tree orders the isolation so each disconnect tests a hypothesis, not a hope.

Symptom presentation

Underproduction shows up as total output below the expected for the irradiance and season. Characterize it:

• One MPPT input or string-level monitor reading low while siblings read normal: a localized string fault, the easy case.
• All strings down proportionally: a system-wide cause (soiling, derating, grid/inverter limit), not a single string.
• Intermittent dips tied to time of day: shading, the array tells you when.
• A hard zero on one input with the rest producing: an open string, blown fuse, tripped rapid-shutdown device, or disconnect.

Quick checks

Get the cheap data before touching DC.

• Read per-string or per-MPPT current and voltage in the inverter or monitoring portal. Compare strings of equal length and orientation; the outlier names itself.
• Compare measured production to expected for the present irradiance (a clear-sky reference or a pyranometer/irradiance app). System-wide shortfall versus single-string shortfall is the first fork.
• Walk the array for obvious causes: soiling, new shading (tree growth, a new vent or antenna), debris, snow, bird fouling.
• Check the inverter event log for ground-fault, arc-fault, isolation, or rapid-shutdown trips that already localize a fault.
• Confirm AC side and grid are not the limiter (a curtailment or voltage-rise trip drops all strings at once, not one).

Isolation tree

Branch first on scope, then sequence within a string fault.

Branch A, system-wide shortfall. Every string is down proportionally. Do not isolate strings. Check soiling, shading across the whole array, inverter derating (high heat, low DC bus), AC/grid limits, or a comms artifact under-reporting. Resolve the global cause; string isolation here finds nothing.

Branch B, one string/MPPT clearly low or zero. This is where sequencing matters. Isolate that string first, it is the named suspect. The order to test it:

Step one, read its open-circuit voltage (Voc). With the inverter safely disconnected and the string isolated at the combiner or disconnect, measure Voc across the string. Compare to expected: expected Voc equals the module Voc times the number of modules in series, temperature-corrected (Voc rises in cold, falls in heat). A near-zero Voc means an open in the string (broken connector, failed module, blown fuse, open conductor). A Voc that is low by roughly one module's worth means one module is bypassed or shorted internally.

Step two, if Voc is open, walk the series path. Sectionalize the string at accessible MC4 junctions, measuring Voc on each half to bisect toward the break (classic binary search: measure the midpoint, the dead half contains the fault, halve again). The open is a connector, a module junction box, or a conductor.

Step three, if Voc is present but current is low, read operating current (Imp/Isc) under load or short-circuit. A string Isc well below its siblings at the same irradiance points to a degraded or shaded module, a high-resistance connection, or a partially failed bypass diode. An IV-curve trace, if you have a tracer, distinguishes shading/mismatch (a stepped curve) from series resistance (a rounded knee) from a current-limited module.

Step four, confirm against a sibling. Swap the suspect string's measured Voc/Isc against an identical healthy string at the same moment. The delta isolates the fault to the suspect, not the inverter input.

Branch C, hard zero on one input with a logged trip. An isolation/ground-fault or rapid-shutdown event already localized the fault. Address the trip cause (insulation fault, water ingress, RSD device) on that string before remeasuring; do not just reset and walk away.

Confirming diagnosis

Tie the fault to a measured number.

• Open string confirmed: Voc near zero on the string, bisection locates a specific connector/module/conductor with the break.
• Bypassed/shorted module confirmed: Voc low by one module's contribution, IV trace or per-module check names the unit.
• High-resistance/degraded module confirmed: Voc near normal but Isc low at matched irradiance, IV knee rounded, sibling delta isolates it.
• System-wide cause confirmed: no single string is an outlier; the global factor (soiling, derate, curtailment) explains the whole shortfall.

Always re-read after the fix in the same irradiance window so the recovery is measured, not assumed.

Remediation

Replace or repair the confirmed component: re-terminate or replace a failed connector, swap a module with a failed cell or bypass diode, correct a high-resistance junction, clear shading or soiling for a global cause, and resolve the insulation/ground-fault path before re-energizing a tripped string. Torque DC terminations to spec and re-verify Voc/Isc per string against expectation after the repair. Reset inverter faults only after the underlying cause is cleared, and confirm full production against irradiance before leaving.

References

• NEC Article 690: Solar Photovoltaic (PV) Systems (DC conductors, disconnects, and 690.12 rapid shutdown).
• IEC 62446-1: Grid-connected PV systems, commissioning, documentation, and inspection (string Voc/Isc verification).
• UL 1741: Inverters, Converters, Controllers and Interconnection System Equipment (inverter and rapid-shutdown context).
• IEC 61829: Photovoltaic array, on-site measurement of current-voltage characteristics (IV-curve method).