Monitoring, SCADA and performance analysis
Monitoring data shows problems before customers call. You will be able to calculate performance ratio, compare actual with expected output, tell a data gap from a real outage and rank alarms by energy lost and safety.
- Core
- Intermediate
- about 14 hours
Written by the GreenHired editorial teamSources last checked 20 November 2026How we research roadmaps
What should the system have produced? That is the first question of every analysis, and it needs sunlight measured in the plane of the array and the module temperature, not the weather forecast. Compare expected with actual first, then look at the shape of the gap. Performance ratio is a screening figure, and it only makes sense once you know the day’s sunlight.
Same system, three days: which one is the problem?
A 5 kWp system means 5 kW of DC rating at standard test conditions. PVPMC gives the IEC 61724 daily AC performance ratio (PR) as the day’s AC energy per kWp divided by the plane-of-array insolation in kWh/m2 (divided by the 1 kW/m2 reference). The three days below use invented figures.
- Sunny day, 5.0 kWh/m2 of sunlight and 20 kWh of AC energy: yield = 20 ÷ 5 = 4.0 kWh/kWp, so PR = 4.0 ÷ 5.0 = 0.80.
- Grey day, 3.5 kWh/m2 and 14 kWh: yield = 14 ÷ 5 = 2.8, so PR = 2.8 ÷ 3.5 = 0.80.
- Odd day, 5.0 kWh/m2 and 15 kWh: yield = 15 ÷ 5 = 3.0, so PR = 3.0 ÷ 5.0 = 0.60.
The grey day gave 30% less energy than the sunny day (6 of 20 kWh), but the sunlight was also 30% lower (1.5 of 5.0), so PR did not move and nothing is wrong. The odd day had the sunny day’s sunlight and 5 kWh less output, which is a real gap. Next questions, in order: was the sensor reading true (clean, in the plane of the array)? Was it a data gap or a real outage? Is one string or one inverter down? Rank what you find by energy lost and by safety.
What a flagged day sends you to next
O&M programmes and inspection routines says what happens on the visit the data triggers. Finding DC and string faults covers what to do once the data points at a string, and inverter faults, repairs and warranty claims covers inverter alarms.
In our sample of 17 postings (UK, US, Germany, India and Australia; how we research roadmaps), 6 wanted inverter configuration, monitoring or SCADA skills, so give the SCADA and data-acquisition part of this topic proper time.
Reading data is safe. Opening equipment to chase what the data suggests is not something to do without authorisation and training.
What you will learn
- What monitoring portals, data loggers and SCADA systems collect, and how often.
- How performance ratio is calculated and why it moves with temperature and season.
- How to normalise output by irradiance and module temperature to judge a day.
- How to separate a communications gap from a real production outage.
- How soiling, shading, clipping and degradation look in data, and how to rank alarms by energy at stake and safety.
The parts of this topic
Portals and string-level monitoringabout 2 hours
What a monitoring portal shows at site, inverter, string or module level, and why the level you watch decides which faults you find.
A portal shows production per site, per inverter and sometimes per string or module. Total output can look fine while one string is dead, because the others carry the day, and a green status light can mean only that the inverter is talking to the portal, not that it is producing as it should. An IEA PVPS report on statistical monitoring of small systems describes comparing metered generation with an expected profile from irradiation data and the system layout. Watch the level where faults hide and compare like with like.
- Improving efficiency of PV systems using statistical performance monitoring IEA PVPS Task 13 · Guide Free PDF, 2017; mostly small systems
Performance ratio and specific yieldabout 2 hours
Performance ratio compares the AC energy you got with what an ideal system of the same rating would give in the sunlight received.
In the IEC 61724 definition used by the Sandia site, daily AC yield (kWh per kWp) is divided by the reference yield, which is plane-of-array insolation divided by 1,000 W/m2. It is sensitive to temperature, so it reads lower in warm months and higher in cold ones. For example, an illustrative drift from 0.82 in spring to 0.77 in midsummer can be normal. Compare performance ratio between sites or years only after checking each irradiance sensor's type, cleanliness and alignment.
- Performance ratio Sandia PV Performance Modeling Collaborative · Guide
Availability and downtimeabout 2 hours
Availability counts the share of possible energy the plant could deliver; good analysis separates data gaps from real outages.
Counting outage hours misleads: a stop at sunny noon costs far more than one at dusk, so energy-weighted availability is fairer. Raising an urgent call-out for a data gap, or missing a real outage hidden inside one, are both costly, so good analysis separates a communications gap, where the plant kept producing but the data stopped, from a true production outage. The open-source RdTools library has an availability analysis that does this and estimates lost energy. A logger fault, for instance, looks like a day of zero output in the portal while the inverter was running.
Expected versus actual outputabout 2 hours
Judge a day by comparing measured output with what measured irradiance and module temperature should have produced.
Expected output comes from measured irradiance in the plane of the array and module temperature, run through a simple model. The free pvlib library shows how, with worked examples. A reference cell or pyranometer is the trusted sensor; satellite data is a fallback with larger error. Example: a grey day with 30% less output than yesterday is not a fault if irradiance fell by the same amount. Check the sensor before blaming the system: a dirty, shaded or differently tilted sensor misleads.
Alarm triage and fault signalsabout 2 hours
Rank alarms by energy at stake and by safety, and learn the four main ways data can reveal a fault.
Alarms outnumber real faults, so triage ranks them: a grid fault stopping a whole plant, an insulation alarm that may mean an earth fault, one string down. One string at 70% of its neighbours for ten days, for instance, outranks a single voltage warning. The IEA PVPS report on advanced algorithms describes four ways to spot faults: electrical signatures, history versus today, predicted versus actual, and comparing similar systems or subsystems. Repeated nuisance alarms breed alarm fatigue, and the dangerous one gets ignored.
Soiling and shading signaturesabout 1 hour
Soiling, shading and inverter clipping each leave a different pattern in the data, which tells you what to check on site.
Soiling often shows as a slow decline that jumps back after rain or a clean, a sawtooth pattern. Shading repeats by time of day and season and is tied to a tree, chimney or neighbouring row. A flat top at the same value is probably clipping at the inverter limit, not dirt. Optimisers do not always help: in one IEA PVPS study of partial shading on urban roofs, string inverters sometimes did as well or better. Look at the time-of-day pattern before calling every dip shade.
- Fact sheet: understanding, measuring and mitigating soiling losses IEA PVPS Tasks 13 and 16 · Guide Free PDF, September 2025
- Performance of partially shaded PV generators with optimised power electronics IEA PVPS Task 13 · Guide
Tracking degradation over yearsabout 1 hour
Modules lose a little output each year; estimating the rate needs clean, normalised data and several seasons.
A degradation rate quoted from one year of noisy data will not hold up in a warranty argument. The Sandia page notes there is no standard procedure for estimating degradation and that results depend on data quality, filtering, aggregation and temperature correction. Year-on-year comparison of normalised data is a favoured method in the open-source RdTools tool, for instance a performance-ratio trend over three years compared season to season.
- PV degradation methods Sandia PV Performance Modeling Collaborative · Guide Links example notebooks and a sample dataset
SCADA and data acquisition basicsabout 2 hours
How loggers, communications and supervisory control collect and move plant data, and what makes the data trustworthy.
A data acquisition system polls inverters, meters and sensors at a set interval, time-stamps the readings and sends them to a server. Larger plants add SCADA for remote alarms and sometimes remote switching. Data quality depends on time stamps, missing-data handling and consistent units; standard data models help. A connected logger left on default passwords, or clocks that drift so two sources never line up, undo the rest. The IEA PVPS digitalisation report also flags cybersecurity as systems become more connected.
Practice task
Take a sample dataset (the Sandia PVPMC degradation page links one, or build a month of hourly values yourself) and calculate daily performance ratio in a spreadsheet. Plot it, mark three days that look wrong and write a one-line hypothesis and the next check for each. This is a data task only; it needs no site access.
Skills you build
Performance ratioExpected versus actual outputAlarm triageAvailability analysisData acquisition basics
Resources for this topic
Every source here is free. Each link opens the provider's own site in a new tab.
- PV Performance Modeling Collaborative Sandia National Laboratories · Guide Modeling guide, datasets and tools
- PVGIS: PV potential tools European Commission Joint Research Centre · Tool
- IEA PVPS Task 13: Reliability and performance of PV systems IEA PVPS · Guide List of free reports on performance and O&M
Live Solar PV technician jobs
Roles whose title matches this career path.
Spot a problem with this topic?
Tell us what is wrong, missing or unclear here. Your note is attached to this topic.
Notes are sent from this page with JavaScript on. Without it, write to us on the contact page.