Solar Monitoring Systems: What to Measure, What to Alarm On, and Who Should Be Watching

Nearly every commercial solar system has some form of monitoring. Usually it is the inverter manufacturer's app, showing today's kilowatt hours and a green tick. That answers one question well: is the system on? It answers the question that costs money badly: is the system producing what it should, given the weather it had?

A monitoring system worth paying for tells you when output is lower than the conditions justify, points to where the loss is, and puts the information in front of someone who will act on it. The last part is where most monitoring fails. The data exists, the alarm fires, and nobody moves.

Three Levels: Inverter Portal, Monitoring System, SCADA

The right level depends on the size of the system and what hangs on its output.

The inverter portal. The manufacturer's cloud platform, fed by the inverter's own measurements. It shows power, energy and fault codes, and it usually comes free. For a small single-inverter system it may be all that is justified. Its limits are structural: it has no irradiance measurement, so it cannot tell a cloudy day from a fault; the inverter's own energy reading is not a calibrated meter; and a portfolio built with mixed inverter brands ends up spread across several portals that nobody checks together.

An independent monitoring system. A data logger on site polls the inverters, an energy meter, irradiance and temperature sensors, and on hybrid systems the battery management system and generator controller. Data goes to a platform that calculates performance, raises alarms and puts every site on one screen regardless of equipment brand. For most commercial systems and multi-site portfolios, this is the right level.

SCADA. Supervisory control and data acquisition adds control to monitoring: a power plant controller that follows set-points from the grid operator for export, reactive power and curtailment, high-resolution data, and integration with substation protection. It belongs on utility-scale and large grid-connected plants, usually because the connection agreement requires it.

The common mistake runs in both directions. A 30 kW clinic system does not need SCADA. A multi-megawatt plant run from an inverter app is losing money it cannot see.

What to Measure

AC energy at a proper meter. A dedicated energy meter at the point of connection gives the number every performance calculation and contract depends on. Inverter-reported energy is useful for diagnosis but is not the figure to settle a dispute with.

Plane-of-array irradiance. A pyranometer or reference cell mounted at the same tilt and orientation as the array. Irradiance is the denominator of every performance calculation, which makes this the most important sensor on site, and one that is easily neglected. A dusty sensor reads low, which makes the plant look better than it is and hides the very losses monitoring exists to find. It needs cleaning on the same schedule as the modules, or more often.

Module temperature. A sensor bonded to the back of a representative module. Without it, the reversible loss from heat cannot be separated from real faults, and every hot month looks like underperformance.

String or input currents. On an inverter with twenty strings, losing one string costs 5 percent of that inverter's output, which disappears inside normal day-to-day weather variation at system level. String-level current, compared across identical strings, finds a blown fuse or a disconnected connector the same day.

Battery state of charge and state of health. On battery systems, read from the battery management system where there is one: state of charge, temperature, depth of discharge each night, cycle count and the trend in state of health. A battery ageing faster than planned shows up here years before it fails.

Generator run hours, starts and fuel. On hybrid systems the diesel is the cost the solar was bought to displace. Run hours, number of starts and fuel level are the measure of whether the hybrid is doing its job. A generator running while the batteries are full means the control logic is wrong, and our article on hybrid solar systems covers how that logic should work.

Grid availability. On grid-connected sites with an unreliable supply, the hours the grid was down. Without this, every grid outage looks like a PV fault.

The measurements that let monitoring tell a fault from the weather. Energy alone cannot. Source: MIMAH engineering practice.

Performance Ratio: Why Energy Alone Lies

Energy output depends on the weather. A low week in the dusty season is normal, and a high week in a clear month can hide a failed string. Judging a system on kilowatt hours alone means judging it on the weather.

The performance ratio removes the weather. It is measured energy divided by what the array's nameplate rating would deliver, with no losses, under the plane-of-array irradiation actually measured. A healthy system holds a fairly stable ratio from month to month once temperature is accounted for. A drop means something changed: soiling, a fault, shading from new growth or a new building, or degradation.

Temperature needs handling deliberately. Modules run hotter in summer and produce less per unit of sunlight, so an uncorrected ratio sags every hot season on a perfectly healthy plant. Either use a temperature-corrected ratio, or compare the measured ratio against the expected ratio from the design model for the same conditions. Over years, the corrected ratio's slope is the system's degradation rate, covered in our guide to solar panel lifespan and degradation.

The IEA PVPS work on analytical monitoring sets out good practice for monitoring hardware and for analysing performance data in this way, including how to understand losses that cannot be measured directly.

IEC 61724-1 Monitoring Classes

The reference standard is IEC 61724-1, which sets out terminology, equipment and methods for monitoring PV system performance and defines classes of monitoring system. The current edition, published in 2021, has two: Class A for high accuracy and Class B for medium accuracy. The 2017 edition also had a Class C, which the 2021 edition eliminated, so a specification still asking for Class C is citing a superseded text.

In practice Class A is typically specified for utility-scale plants and large commercial systems where performance guarantees carry money, and Class B for smaller commercial installations. The class is not a matter of buying the right sensors. It depends on the equipment, the calibration and cleaning regime, and the inspection procedures together. A Class A pyranometer that nobody cleans does not make a Class A system.

Two points in the standard are worth reading before writing a specification. It says the monitoring system should be matched to the system's size and the user's requirements, with larger systems having more measurement points and more accurate sensors than smaller, lower-cost ones. And it distinguishes two purposes: comparing performance with design expectations and guarantees, which needs consistent system-level data, and detecting and localising faults, which needs finer resolution within the system. Most owners need both, and an inverter portal without irradiance data does neither properly. Its main clauses are written for grid-connected systems without storage or backup sources, with only an annex giving some detail for systems that have them, so hybrid monitoring requirements need writing into the specification explicitly.

Alarms That Matter, and Alarm Fatigue

A platform that emails every inverter warning to everyone gets filtered to a folder within a month, and then the alarm that matters is missed with the others. Alarms should be graded by consequence and routed to someone who can act.

Act now. Insulation or earth faults reported by the inverter, which are safety issues. Battery over-temperature or a battery management system fault. A site producing nothing in full daylight. A generator that failed to start when called.

Act today. Underperformance against irradiance sustained over hours, not a single interval. A string or inverter input well below identical neighbours. A generator running while the batteries are full. Low battery state of charge at dawn.

Review weekly. Communication dropouts that recovered. Slow drift in performance ratio. Rising generator hours. A falling trend in battery state of health.

Thresholds should be set from a few weeks of real data after commissioning, not from defaults, and reviewed monthly for the first year. The rule that keeps alarm fatigue away is simple: every alarm has an owner and an action. An alarm nobody acts on should be re-thresholded or removed, because its only effect is teaching people to ignore the others.

Alarms graded by consequence. Every alarm needs an owner and an action, or it teaches people to ignore the rest. Source: MIMAH engineering practice.

Keeping Data Flowing From Remote Sites

Monitoring in Sudan and much of rural Nigeria has a problem that monitoring brochures ignore: the connection.

Cellular first, satellite where it must. Most remote sites communicate over GSM or other cellular data. Where coverage is weak or absent, satellite links work at higher cost. The choice should follow a signal survey at the actual equipment location, not the coverage map.

Local buffering. The data logger should store data on site for long enough to ride through a communication outage and send it when the link returns. Specify the buffer duration. A logger without one turns every network outage into a permanent gap.

Power the monitoring from the battery side. A logger and router that shut down when the grid or the generator does lose the data for exactly the periods you most need to see.

SIM and account housekeeping. Expired data bundles, lapsed SIM registrations and changed passwords are a common and entirely avoidable reason for a remote site to go dark. Someone must own them.

Heat and security. Loggers and routers in unventilated enclosures on hot sites fail early. Change default passwords and do not expose site equipment directly to the internet.

The O&M contract should also define how missing data is treated in performance and availability calculations, before the first gap appears. MIMAH's 2025 programme in Sudan installed about 1.39 MW across 110 sites, systems of 7.5 to 50 kW, and at that scale the monitoring has to be affordable per site while still telling a technician which of 110 sites needs a visit this week.

Who Watches, and What They Do

Monitoring with no response attached is a well-documented record of losses. The O&M contract is where response gets defined: who receives each alarm tier, the response time for each, how much can be diagnosed remotely before a visit, and what gets reported to the owner. Our solar O&M contract guide covers those clauses, and the pattern described in why solar systems fail early is often one of warnings nobody acted on.

Owners and technicians need different views of the same data.

The owner's dashboard. Monthly energy against expected, performance ratio, availability, diesel displaced on hybrids, and open issues with their status. It should answer whether the asset is earning what it should, and every deviation in the monthly report should come with an explanation.

The technician's dashboard. Live alarms by priority, string and inverter data, fault codes, battery and generator detail, and the history of each site. It should answer what is wrong, where, and what to bring.

Same data, two audiences. The owner needs to know whether the asset is earning; the technician needs to know what to fix and what to bring. Source: MIMAH engineering practice.

Monitoring Starts at Commissioning

A monitoring system is part of the plant and should be commissioned with it: sensors checked for mounting and orientation, the irradiance sensor verified, the meter reading compared with the commissioning measurements, alarm routes tested end to end, and a baseline performance ratio recorded for the first weeks. Our guide to solar PV commissioning covers the tests that baseline should sit beside.

MIMAH's renewable energy team provides remote monitoring for solar and hybrid systems as part of its renewable services, alongside preventive maintenance and performance audits. If your system reports numbers that nobody is turning into action, get in touch.