Solar PV O&M field guide: utility-scale plant maintenance and monitoring

What solar PV O&M is, and why the plant loses money quietly

Solar PV operations and maintenance is the discipline of keeping a photovoltaic plant producing the energy the financial model assumed it would, across a service life of 25 to 30 years. A solar array has no moving sign that it is failing. Unlike a pump that screams or a building that leaks, a soiled array, a tripped string, or a faulty inverter simply makes less power while the site looks completely normal. The lights stay on. The loss is real and it is invisible.

That is the whole problem O&M exists to solve. The work splits into two halves that depend on each other. The first half is watching the production data closely enough to notice when the plant is making less than it should, which is harder than it sounds because the weather changes the output every hour. The second half is the physical maintenance that recovers the loss: cleaning soiled modules, servicing inverters, scanning for dead modules and strings, and re-torquing connections that thermal cycling has worked loose.

All of it happens around one fact that shapes every procedure on the site. A PV array is energized whenever the sun is up, and you cannot turn off the sun. The DC side stays live in daylight even with the breakers open. This guide covers the O&M side of solar. For the install and wiring rules see the photovoltaic system wiring guide, and for the storage that increasingly sits next to these plants see the battery energy storage guide.

Why a soiled or failing plant still looks fine

The single most expensive truth in solar is that underperformance is invisible without monitoring. A diesel generator that loses a cylinder runs rough and you hear it. A solar plant that loses a string, a combiner, or a whole inverter keeps sitting in the sun looking identical to the day it was commissioned. The only evidence is on the production curve, and only if someone is reading it against what the plant should have made for that day's irradiance.

Walk a site that has gone a year without real monitoring and the losses stack up unseen. A few percent off the top for soiling that never got cleaned. A central inverter that has been derating on a thermal fault since spring. A handful of strings dark because a combiner fuse blew and nobody knew. Each one bleeds revenue every sunny day, and none of them throws an alarm a passerby would notice.

This is the framing that should drive the program: you either monitor the plant or it bleeds. The owner who skips monitoring is not saving money on O&M. They are choosing not to find out how much they are losing. The loss does not stop because nobody is watching. It just goes unmeasured until a year-end production shortfall forces the question, by which point the money is already gone.

Performance monitoring: the eyes on the plant

Performance monitoring is the part of O&M that makes the invisible loss visible. The plant reports its production through meters and the inverters report through their data interfaces, and a monitoring platform compares what the plant actually made against what it should have made given the irradiance, the temperature, and the plant's rated capacity. The gap between expected and actual is the alarm that no walkthrough will ever give you.

Good monitoring catches the drop fast. A string that goes dark shows up as a step down in one inverter's output relative to its neighbors. A soiling trend shows up as a slow daily divergence between expected and actual that recovers after rain. An inverter fault shows up as a derate or a zero. The expected-versus-actual comparison is the heart of it, because raw production alone tells you nothing without the irradiance to judge it against.

The field side of monitoring is closing the loop. An alert is only worth something if it turns into a dispatched technician, a work order, and a recorded fix. A field tool like FieldOS is where the alarm becomes a tracked job: the fault, the location, the technician assigned, the parts used, and the production recovered, all in one record instead of scattered across texts and a spreadsheet. The platform sees the loss. The field system makes sure somebody acts on it.

What are performance ratio (PR) and availability?

Performance ratio (PR) is the plant's actual energy output divided by the energy it should have produced from the measured irradiance and its rated capacity, expressed as a fraction or percent. It is the headline measure of how well a PV plant converts available sunlight into delivered energy, and it strips the weather out so you are comparing the plant to itself, not to a sunny day. The international reference for how to measure it is IEC 61724.

Availability is the other half of the scorecard. It measures the fraction of time, or of producible energy, that the plant was actually able to operate rather than sitting down on a fault. The two answer different questions. PR asks how efficiently the plant ran when it ran. Availability asks how much of the time it was up. A plant can post a healthy PR and still bleed money if a central inverter is down two weeks a quarter, which is why both numbers belong on every report.

Watch the trend, not the single day. A PR that drifts down month over month is the underperformance flag, and the cause is usually soiling, degradation, or a fault that has not been chased. These figures are the common industry KPIs, but the exact definitions, the measurement boundary, and the guarantee thresholds vary by contract and standard, so confirm them against the project agreement and the IEC reference before you hold anyone to a number.

Preventive, corrective, and condition-based work

O&M work comes in three modes, and a healthy plant runs a mix of all three. Preventive maintenance is scheduled: the cleaning, the vegetation cuts, the connection torque checks, the inverter filter and fan service, the annual thermography scan. You do it on a calendar before anything has failed, because most of these tasks are cheaper as prevention than as a fault response.

Corrective maintenance is the fix after something breaks: replace the failed inverter, swap the blown combiner fuse, change out the cracked module, repair the tracker that stopped following. It is unavoidable, and the measure of a good program is how fast corrective work gets dispatched and closed, because every day a fault sits open is production lost.

Condition-based maintenance is the modern third leg, where the data itself triggers the work. The monitoring platform flags a soiling trend that has crossed the economic threshold, or a string current that has drifted, or an inverter temperature creeping up, and you act on the signal rather than on the calendar or on the failure. The art is the blend. Over-schedule and you spend money cleaning a clean array. Wait for failure and you lose production you could have saved. Let the production data tell you where to spend the maintenance dollar.

How often should solar panels be cleaned?

Cleaning frequency is set by economics, not by a calendar, and the right interval ranges from never to weekly depending on the site. Soiling is the buildup of dust, pollen, bird droppings, pollutants, and snow on the module glass, and it blocks light before it reaches the cell. Industry surveys put average soiling losses around 4 to 7 percent of annual energy, but in dusty or desert climates the loss runs far higher. NREL field work has measured desert dust cutting yield by as much as 40 percent, and dust deposition studies report 20 to 50 percent losses in the worst conditions.

Cleaning pays when the recovered energy revenue beats the cleaning cost, and that math depends entirely on local soiling rate, electricity price, and how much rain washes the array for free. NREL O&M best practice has documented cleaning costs in the range of roughly $0.25 per square meter for large systems up to about $1 per square meter for small ones, which gives you the cost side to weigh against the recoverable loss. A soiling station, a pair of reference modules where one is kept clean and one is left to soil, measures the actual loss so you clean on data instead of on a guess.

On utility-scale sites the cleaning method is its own decision. Robotic cleaning has matured to where waterless robots can run scheduled passes across long rows, and recovery on heavily soiled arrays can be substantial. Water trucks and crews still do the work on many sites. Match the method to the water availability, the dust load, and the labor cost. Soiling is the most recoverable loss on the plant, and it is also the one owners most often ignore because the array still looks like it is working.

Why is the inverter the most common source of downtime?

The inverter is the part that fails most and takes the most production down with it. Analysis of fleet data has attributed on the order of 80 percent of PV plant downtime and failures to the inverter, which makes it the component your O&M program lives or dies on. It is the box doing the hardest work, switching high power continuously and running hot, and heat is what kills it. Overheating from clogged air filters, failed cooling fans, and blocked heat sinks is the leading cause of premature inverter failure, and most of that is preventable with service.

Preventive inverter work is unglamorous and high-payback: clean or replace air filters on the manufacturer's interval, which can be every six months in dusty conditions, verify the cooling fans, clear the heat sinks, keep the firmware current, and read the fault and event logs for the warnings that precede a hard trip. The faults themselves fall into electrical, thermal, and communication buckets, and many clear with a firmware update or a filter change rather than a replacement.

Central versus string changes the risk shape. A central inverter failure can take a large block of the plant offline at once, so the downtime per event is severe and the spare is a long-lead, high-value item. String inverters spread the risk, so a failure only darkens a section, but you have many more units to service. Either way, track the warranty term and the spares position, because an inverter past warranty with no spare on the shelf is a multi-week outage waiting to happen. The exact service intervals and fault meanings are manufacturer-specific, so the inverter manual governs.

Thermography and IV-curve tracing: finding the invisible faults

Thermography and IV-curve tracing are the two diagnostics that find the faults the production meter can see but not locate. Infrared thermography, usually flown by drone across a large site, images the heat signature of every module. A dead module reads cold, a hot spot reads as a bright point or a heated cell, a bypassed substring shows a banded pattern, and a failed combiner or a bad connector lights up where the heat concentrates. One flight can scan a plant that would take a crew weeks to inspect module by module, and it turns a vague PR shortfall into a map of exactly which modules and strings to visit.

IV-curve tracing is the electrical complement. A curve tracer sweeps a module or string through its current-voltage characteristic and the shape of the curve tells you the health: a notch or step points to a shaded or cracked cell or a bad bypass diode, a depressed curve points to series resistance or degradation, a low voltage points to a cell or connection problem. Where thermography says which string looks wrong from the air, the IV trace confirms what is wrong electrically on the ground.

Run thermography on a schedule, commonly once a year and again after any major weather event, and trace the suspect strings the scan flags. These tools find the losses that are otherwise invisible until they show up as a number you cannot explain. Pair both with electroluminescence imaging when you need to see cell cracks the IR misses. Interpret the results against the module datasheet and the manufacturer's guidance, because what counts as an out-of-spec curve depends on the module.

Hot spots, PID, and module degradation

Modules fail in a handful of recognizable ways, and naming the mechanism tells you whether it is recoverable or a replacement. A hot spot is a single cell or area running far hotter than the rest, usually from a crack, a shaded cell forced to dissipate power, a short-circuited bypass diode, or a soldering defect. Left alone it cooks the encapsulant and can become a fire risk, so a hot spot the thermography flags gets traced and the module pulled.

Potential-induced degradation, PID, is a string-level loss that comes from the high system voltage driving ion migration through the encapsulant between the cells and the grounded frame, especially in heat and humidity. It can drop an affected string's output sharply, on the order of 10 to 30 percent within a year, and depending on the system it can sometimes be partially recovered with anti-PID equipment that reverses the voltage stress overnight. LeTID, light and elevated temperature induced degradation, is a separate mechanism that develops later in life and adds a few percent loss in hot climates on certain cell types.

Underneath all of it is normal degradation. A quality module loses roughly 0.4 to 0.5 percent of output per year over its 25-to-30-year life, with a larger drop in the first year, and the warranty is written around that curve. The job of O&M is to separate the slow, expected degradation from the fast, abnormal failures, because the first is priced into the model and the second is a warranty claim or a replacement you act on now. The degradation figures and warranty thresholds are module-specific, so the datasheet and warranty document control the call.

Combiners, fuses, and MC4 connectors

The DC string path is where small failures hide and where fires start. Combiner boxes gather many strings onto a single feed, each string through a fuse, and a blown fuse silently takes that string offline. Nothing alarms at the combiner level on many designs, so a handful of dark strings can sit unnoticed for months. Open the combiners on the preventive schedule, check the fuses, and watch the per-string currents in the monitoring data where the plant reports them.

Connectors are the part that punches above its size. The MC4 is the standard DC connector across the industry, and it is also the most common single failure and fire point on a PV plant. The failure mode is a bad mate: a poorly crimped contact, a connector seated against a different brand, or a joint that worked loose, any of which builds resistance, which builds heat, which can arc. A DC arc does not self-extinguish the way an AC arc does, so a degrading connector is a real ignition source, not a nuisance.

The discipline here is plain. Use connectors of one brand, crimped with the matching tool to the manufacturer's spec, and never mate two different brands even when they physically click together. On thermography, a hot connector is a finding you act on, not a curiosity. Most string-level losses and most DC fires on these plants trace back to the combiner and the connector, so that is where the inspection attention earns its keep. The connector ratings and crimp specs are manufacturer-specific and listed, so follow the connector manufacturer's instructions.

Re-torque and the hot connection

Electrical connections loosen over time and the cause is thermal cycling. Every day the conductors and lugs heat under load and cool at night, and that expansion and contraction works a connection loose over months and years. A loose connection has resistance, resistance makes heat at the joint, and heat accelerates the loosening, which is the runaway that ends in a discolored lug, a melted terminal, or a fire.

The preventive answer is to check and re-torque the electrical connections on a schedule: the combiner terminations, the inverter DC and AC lugs, the disconnect and panel connections, torqued to the value on the equipment or in the manufacturer's table, not by feel. Thermography catches the ones already running hot, but the torque check catches them before they get there. Mark torqued connections so the next technician can see what has been verified.

The electrical joints are not the only ones that loosen. The mechanical hardware matters too. Racking bolts, module clamps, and tracker fasteners take wind and thermal loading, and a backed-out clamp lets a module move or lift. The torque on those is a structural item the racking manufacturer specifies. Check the connection torque on the schedule, because the hot connection is one of the few failures that gives you warning, by smell and discoloration, before it gives you a fire.

Trackers: the moving parts of a utility-scale plant

Single-axis trackers add production by following the sun east to west, and they add the only real mechanical maintenance load on a utility-scale plant. Each tracker row has a motor, a drive, bearings, structural members, and a controller, and any of them can fail in a way that stops the row from tracking or freezes it at a bad angle. A tracker stuck flat or stuck at the wrong tilt is a production loss that, like a soiled array, looks fine from the gate.

The maintenance is mechanical and routine: lubricate the motors and drives on the manufacturer's interval, which is often around every six months, inspect the gears and bearings at least annually, check the structural fasteners for backout, and verify the controller and its position feedback. Drive systems on quality trackers are rated for the 20-plus-year plant life under standards like IEC 62817, but only if they are serviced.

The wind stow function is the safety-critical part. When wind exceeds a threshold, commonly in the range of 18 to 22 meters per second, the controller drives every row to a low-angle stow position to cut the projected area and the uplift force that could tear the array apart. A stow system that fails to trigger is a structural risk, so it gets verified, not assumed. One caution worth carrying: independent field studies have found real tracker availability running well below the 99 percent figures often claimed, in some fleets between roughly 66 and 88 percent, and trackers are frequently left out of performance guarantees. Monitor tracker uptime as its own number, because nobody else may be. Service intervals and stow thresholds are tracker-specific, so the manufacturer's manual governs.

Vegetation and erosion: keeping the site under control

Vegetation is a production and a site-integrity problem at once. Grass and brush growing up under and between the rows shades the lowest cells, and a shaded cell on a string can drag the whole string down out of proportion to the area it covers. Left to grow, vegetation also becomes a fire fuel load and blocks access to the rows for the rest of the maintenance.

The control options are mowing, string trimming around the posts and equipment, targeted herbicide where allowed, and on a growing number of sites, sheep. Grazing sheep keep the growth down without machinery near the array, and the practice, sometimes called solar grazing, has become a normal vegetation strategy on utility-scale land. Pick the method for the site, the local rules, and the cost.

Erosion is the slower site failure. The same large graded site that holds the array sheds water, and if the stormwater controls and ground cover are not maintained, channels form, posts undermine, and the racking geometry shifts. Keep the vegetation cover that holds the soil, keep the swales and culverts clear, and walk the site after heavy rain. The array is only as stable as the ground it stands on.

Why is solar DC dangerous?

Solar DC is dangerous because the array is energized whenever there is daylight and you cannot turn it off. This is the fact that makes PV different from almost everything else an electrician works on. You can open every breaker and disconnect on the plant, and the modules outside are still generating, still pushing DC voltage onto the conductors between the array and the inverter. The sun is the source, and there is no switch for the sun.

That means large parts of the DC system cannot be put into a verified dead state during daylight. Rapid shutdown, required by the NEC at 690.12, de-energizes the conductors inside and near the building to a safe level for first responders, but it does not de-energize the array itself out in the field, and it is a life-safety function, not a maintenance lockout. A string can hold hundreds of volts DC in full sun with every disconnect open.

So the rule on the DC side is simple and absolute: treat the DC as live whenever the sun is up, because it is. Verify dead with a meter you have proven on a known source, understand the limits of what you can actually de-energize, and do not assume an open disconnect made the array safe. DC at these voltages does not let go the way people expect, and the hazard is there on the brightest, most ordinary day. Follow the safety requirements in NFPA 70E for DC work and the manufacturer's procedures. This is the section to be blunt about: if you would not grab it, do not assume it is dead because a switch is open.

DC arc-flash and rapid shutdown

A DC arc is more stubborn than an AC arc, and that is the heart of the PV arc-flash hazard. AC current crosses zero 120 times a second, and each zero-crossing gives an arc a chance to extinguish. DC has no zero-crossing, so once a DC arc strikes it tends to sustain itself until something opens the circuit or the fault clears. That is why a degrading connector or a loose lug on the DC side is a genuine ignition source, and why DC arc-flash energy is treated as its own hazard.

NFPA 70E covers electrical safety in the workplace and addresses DC hazards alongside AC, with the standard's approach to DC arc-flash assessment evolving as the research matures. The honest state of the field is that DC arc-flash modeling is less settled than AC, so the conservative approach in the standard plus a real risk assessment is how it gets handled. The work is done by qualified persons, in the rated PPE, following an energized-work justification when the DC genuinely cannot be de-energized.

Rapid shutdown under NEC 690.12 reduces the shock hazard for firefighters and responders by collapsing the voltage on the conductors in and on the building when the system is shut down. Know what it protects and what it does not. It is a responder-safety feature for the building conductors, not a guarantee that the array or the DC field wiring is safe to handle. The arc-flash boundary, the PPE category, and the shutdown details depend on the system and the adopted code edition, so confirm them against NFPA 70E, NEC Article 690, and the equipment documentation.

SCADA, the monitoring platform, and the data

On a utility-scale plant the SCADA system and the monitoring platform are how a handful of people operate hundreds of acres of equipment. SCADA gathers the data from the inverters, the meters, the trackers, the weather station, and the combiners, gives the operator remote visibility and control, and feeds the analytics that turn raw points into alarms and performance numbers. Without it, a plant this size is unmanageable, because no crew can walk every string every day.

The data flows through industry interfaces so equipment from different makers reports in a common form, with the SunSpec specifications widely used for that interoperability. The platform's value is in the analytics on top: the expected-versus-actual comparison, the inverter-to-inverter benchmarking that flags the one box that has slipped, the soiling trend, and the alarm prioritization that tells the operator which of fifty alerts is the one costing real money this hour.

The platform spots the problem. The field still has to fix it, and that handoff is where production gets recovered or lost. A field tool like FieldOS carries the alarm into a dispatched work order with the location, the asset, the technician, the parts, and the close-out, so the SCADA alert and the physical repair live in one chain of record instead of dying in an inbox. Define which alarms auto-generate a job and which get triaged, so the operator is acting on the costly ones, not drowning in nuisance trips.

Spares, warranty, and liquidated damages

Spares strategy is what decides whether a failure is a day of downtime or a month. The long-lead items are the ones that hurt: a central inverter or a major power module can run weeks or longer to source, and a string of obsolete modules may be impossible to match. The asset that has critical spares on the shelf rides out a failure. The one that orders after the fault eats the full lead time in lost production, every sunny day of it.

Warranty is the other lever, and it only works if you track it and claim it. Modules carry a product warranty and a separate performance warranty written around the degradation curve, inverters carry their own term, and the work of O&M is to catch a failure that falls inside the term and file the claim before it lapses. A module degrading faster than the warranty curve allows is a claim. An inverter that fails in its term is a claim. Miss the documentation and you pay for what the manufacturer owed you.

On the contract side, the production guarantee and liquidated damages are how risk gets allocated. O&M agreements commonly carry an availability guarantee, often near 99 percent, and a performance-ratio guarantee, and when the plant falls short the agreement defines the liquidated damages that compensate the owner for the lost revenue. Those thresholds and the LD calculation are negotiated and vary by deal, so the contract documents control what is owed. Whether you are the owner or the provider, the numbers only hold up if the production and availability data behind them are clean.

The commissioning baseline

Everything O&M judges the plant against starts at commissioning. The capacity test and the initial performance ratio establish the baseline, the proof that the plant delivers the rated output under reference conditions on day one. Without that baseline you have nothing to measure degradation or underperformance against later, because you never established what good looked like.

At handover from the EPC, the O&M provider should receive the full record: the as-built drawings, the equipment datasheets and serial numbers, the warranty documents, the commissioning and capacity-test results, and the baseline production data, with a joint walk to verify plant condition. That package is the reference the whole maintenance program reads from, and a thin handover haunts the asset for its life.

Treat the baseline as a living reference, not a filed document. When you re-test or scan the plant, you are comparing to the commissioning numbers to separate normal degradation from a developing fault. The capacity-test method, the reference conditions, and the acceptance criteria are defined by the contract and the applicable test standard, so confirm them there rather than assuming a single figure applies to every plant.

Grounding, lightning, and surge protection

A solar field is a large, flat, exposed conductor sitting in the open, which makes grounding and lightning protection part of the maintenance, not just the install. The array structure and equipment are bonded and grounded so fault current has a path and so the structure does not float at a dangerous potential. Over years, ground connections corrode and loosen like any other, so the grounding and bonding gets inspected and tested on the same cycle as the rest of the connections.

Lightning and surge are the exposure that comes with the open site. A direct strike or a nearby strike couples a surge onto the array and the conductors, and that surge is a leading killer of inverters and monitoring electronics. Surge protective devices, SPDs, are installed on the DC and AC sides and at the equipment to clamp those transients, and an SPD that has taken hits eventually fails and stops protecting. Check the SPD status indicators on the schedule and replace the spent ones, because a dead SPD looks the same as a live one until the next surge proves otherwise.

The grounding and bonding rules for PV live in NEC Article 690 and the general grounding articles, and they tie into the wiring side covered in the photovoltaic system wiring guide. The specific SPD ratings, the grounding electrode arrangement, and the lightning protection scheme depend on the site and the design, so the engineering documents and the adopted code edition govern.

The O&M contract: scope, SLA, and reporting

The O&M contract is the document that decides whether the plant is actually maintained or just nominally covered. The scope defines what the provider does: the preventive schedule and its frequencies, the corrective response, the cleaning, the vegetation, the thermography, and what is carved out as a separate cost. A vague scope is where the cheap bid hides, because the work that is not named is the work that does not happen.

The service level agreement sets the response. A real SLA obligates the provider to react to alarms within a defined time, seven days a week, with price adjustments or damages when the response time is missed, because a fast alarm is worthless if nobody is bound to act on it. The availability and performance guarantees, with their liquidated damages, sit alongside the SLA as the teeth.

Reporting is the part owners underweight and regret. The contract should require regular production and performance reporting, the PR and availability against guarantee, the loss accounting that says where the missing energy went, and the maintenance record of what was done. That reporting is how an owner who is not on site knows whether the asset is healthy. The specific SLA times, guarantee thresholds, and LD formulas are negotiated per deal, so the executed contract controls, not any rule of thumb.

Production reporting and loss accounting

Production reporting is where the plant proves it earned what the model promised, and the most useful report does loss accounting: it does not just show the energy made, it explains the energy not made. So much was lost to soiling, so much to inverter downtime, so much to grid curtailment, so much to degradation. That breakdown turns a bare shortfall into a list of fixable line items and a case for where to spend the next maintenance dollar.

The report ties the field work back to the money. When a string was dark for three weeks, the report shows the production lost and the date the repair recovered it, which is how an owner sees the value of the O&M and how a provider defends the work it did. The PR and availability against the guarantee anchor it, and the loss categories explain the gap.

Pulling this together by hand from inverter exports and a maintenance spreadsheet is slow and error-prone. A field tool like FieldOS that holds the work orders, the repairs, the parts, and the close-out dates lets the loss accounting connect each unit of lost production to the fault that caused it and the fix that ended it, so the report writes from records instead of memory.

The records that hold the asset's value

A PV plant is sold and refinanced on its records as much as on its hardware, and a plant with thin records sells at a discount because the buyer cannot see what they are getting. The maintenance history, the monitoring data, the thermography scans year over year, the inverter service logs, and the warranty claim trail are the evidence that the asset was cared for and that the production it shows is real.

Each record category does specific work later. The monitoring history shows the production trend and the degradation curve. The PM logs show the schedule was kept. The thermography sets show how the module population aged. The inverter service records show the critical component was maintained. The warranty file shows what is still claimable. Lose any of them and you lose the ability to prove the corresponding part of the story.

Keep the records in one system rather than scattered across email, paper, and a technician's truck. A field tool like FieldOS that captures each visit, each scan, each part, and each close-out builds the asset history as a byproduct of doing the work, so when the plant changes hands or a warranty fight comes up, the record already exists instead of being reconstructed after the fact.

What to document

The record is what lets the next person, or the next owner, see what the plant has actually been through. Capture the production and performance data continuously, the preventive work as it is done, the corrective fixes with the production they recovered, and the diagnostic scans on their schedule. The goal is that any question about the plant's health has an answer in the file, not in someone's memory.

Common mistakes

• Running with no real performance monitoring, so the losses stay invisible until a year-end shortfall.
• Ignoring soiling because the array still looks like it is working.
• Neglecting the inverters, letting clogged filters and failed fans overheat the most failure-prone component.
• Skipping thermography, so dead modules and hot connectors go unfound until they show up as numbers nobody can explain.
• Loose or arcing DC connectors from mismatched brands or bad crimps, the most common fire point on the plant.
• Working the DC side as if an open disconnect made it dead, when the array is live in any daylight.
• Carrying no spares for long-lead inverters and modules, turning a failure into weeks of lost production.
• Leaving trackers out of the monitoring, so a stuck row loses output unseen.

Field checklist

Standards and references

The installation and safety framework for PV in the United States is NEC Article 690, NFPA 70, which covers the wiring, the disconnects, the grounding, and rapid shutdown at 690.12. Electrical safety for the work itself, including the DC arc-flash hazard and the energized-work rules, is NFPA 70E. Those two govern how the plant is built and how it is safe to work on, and the adopted code edition and local amendments control which version applies.

Performance measurement leans on IEC 61724 for how performance ratio and monitoring are defined and computed, and the SunSpec specifications for how the equipment reports data interoperably. For O&M practice itself, the NREL Best Practices for Operation and Maintenance of Photovoltaic and Energy Storage Systems is the widely referenced guidance on cleaning economics, maintenance scope, and program design, and IEA PVPS publishes related international best practice.

Three things outrank any general reference for the specific call. The module, inverter, tracker, and connector manufacturers' manuals govern service intervals, torque values, fault meanings, and stow thresholds. The project's O&M contract and PPA govern the guarantees, the SLA, and the liquidated damages. And the engineering documents govern the grounding, surge, and protection scheme. Cite the standard that controls the point, and let the manufacturer's instruction and the contract override any rule of thumb. Confirm section numbers and editions against the documents in force before you put them on a submittal.

Units and terms

Solar O&M carries its own vocabulary, and the same idea reads differently across a SCADA screen, a manufacturer sheet, and a contract. These are the terms that recur through the work.

Solar PV O&M

Operations and maintenance of a photovoltaic plant: monitoring production and doing the preventive and corrective work that keeps it producing over a 25-to-30-year life

Performance ratio (PR)

Actual energy output divided by the energy expected from measured irradiance and rated capacity, the headline efficiency metric, referenced to IEC 61724

Availability

The fraction of time or producible energy the plant was able to operate rather than sitting on a fault

Soiling

Loss from dust, pollen, droppings, pollutants, or snow on the module glass blocking light, commonly 4 to 7 percent of annual energy and far higher in dusty climates

Thermography / IV-curve

IR imaging that finds hot spots and dead modules and strings, and current-voltage tracing that confirms module or string health electrically

PID / degradation

Potential-induced degradation, a voltage-driven string loss, against normal aging of roughly 0.4 to 0.5 percent per year over the module's life

MC4 connector

The standard PV DC connector, and the most common single failure and fire point when crimped poorly or mated across brands

Rapid shutdown

NEC 690.12 function that collapses voltage on the building conductors for responder safety, which does not de-energize the array in the field

Tracker

A single- or dual-axis mount that follows the sun to add production, with a motor, drive, bearings, and a controller that wind-stows in high wind

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is NEC 690 rapid shutdown?
• What is a BESS?
• What is predictive maintenance?
• What is electrical submetering?
• What is elevator modernization?