Solar Farm String and Combiner Monitoring

Back to Blog

Why Comparing String Currents Works

A solar farm wires many panels in series into strings, and many strings into a combiner box that feeds an inverter input. The physics that makes string monitoring so powerful is simple: strings of the same design, at the same tilt and orientation, under the same sun, should produce nearly the same current. When one string reads noticeably lower than its siblings, the difference is the fault, and the identity of the string is the location.

This comparison is more sensitive than anything the inverter can offer alone, because the inverter sees only the summed current of all its strings. A single dead string among twenty is a small percentage of the inverter's total and can hide inside normal variation, but it is glaring when you compare that string against the other nineteen. This is the same peer-comparison principle that makes per-input analysis valuable across PV, and it complements the aggregate view discussed in the wider guide on what to monitor at a solar farm.

What Combiner Monitoring Actually Catches

The failures string monitoring localizes are the bread and butter of solar O and M. A blown string fuse drops that string to zero and is invisible from the meter but obvious in a string comparison. Progressive soiling shows as a slowly widening gap between the dirtiest strings and the cleanest, informing when a cleaning cycle pays for itself. Partial shading from vegetation growth appears as a time-of-day pattern on specific strings. A failed module or a degraded connector shows as a persistent low reading that does not track the sun.

Combiner boxes increasingly carry their own monitoring electronics that measure each input current, and sometimes fuse status, box temperature, and surge-protection-device health. Box temperature and SPD status matter because combiners sit outdoors carrying real current, and a hot combiner or a failed surge device is both a fire-risk and a reliability concern. These points extend string monitoring from purely a production tool into a safety-adjacent one, so any thermal or SPD alarm from a combiner deserves a field visit and should be handled per the site's electrical safety procedures by qualified personnel.

Prioritizing string faults is a likelihood game. Because they are common and each one is a fixed percentage of an inverter's output, a plant that trends string currents daily and acts on the widening outliers recovers energy that a meter-only operation simply never sees, a discipline aligned with formal condition monitoring practice.

Normalizing Before You Compare

Peer comparison only works between true peers. Before trusting a string outlier, confirm the strings being compared share the same module count, the same orientation, and the same tracker group - a string with fewer modules in series or a different tilt will read differently forever, and flagging it daily just trains people to ignore the alarm list. Group strings by design first, then compare within the group.

The second normalization is against conditions. Absolute current thresholds fail because every passing cloud crosses them; the robust metric is each string's current as a fraction of its peer group's median at the same moment. A ratio like that is largely self-correcting for irradiance, season, and time of day, because whatever the sun is doing, it is doing it to the whole group. Practical implementations also gate the comparison: evaluate only above a site-chosen irradiance or inverter-output floor, because at dawn, dusk, and under heavy cloud the currents are tiny and the ratios are dominated by noise.

A Worked Symbolic Example

Call the median current of a combiner's strings I-med, and express each string as a ratio r of I-med. The ratio's behavior over a day is the diagnosis. A healthy string holds r near 1 through the whole solar day. A string with r at essentially zero from sunrise to sunset is an open circuit - a blown fuse, a failed connector, or an open in the wiring - and is the highest-value find because it is a full string of lost production with a usually cheap fix. A string whose r sits persistently below 1 by a roughly constant fraction all day points to soiling or module degradation on that string. A string whose r dips only during particular hours and recovers is shading, and the hours tell you where to look for the obstruction.

The alarm logic falls out of the same picture: require the deviation to persist across consecutive evaluation periods within the gated daytime window before alarming, and rank strings by how far and how long their ratio has departed from the group. That ranking, trended day over day, is effectively a work-order queue sorted by recoverable energy.

Getting String Data Into the SCADA Layer

Monitored combiner boxes typically carry a small measurement board reporting per-input current, and often fuse status and box temperature, over a serial fieldbus - commonly Modbus RTU on an RS-485 multidrop - back to a data concentrator or gateway. On plants built with string inverters instead of central inverters, much of the same granularity comes free: a string inverter measures its own DC inputs, so per-input data can be read from the inverter fleet without separate combiner instrumentation, alongside the AC-side quantities covered in solar farm inverter monitoring points.

Whichever path the data takes, naming discipline decides whether the dataset is usable at scale. A utility plant has thousands of strings, and a fault report is only actionable if the tag name encodes the physical location - block, inverter, combiner, input - so a technician can drive to the right box without a lookup table. Establish that convention before commissioning, because renaming thousands of tags afterward is a project nobody schedules.

Frequently Asked Questions