Recent research suggests smarter software could change how Australian households understand their solar systems — from production to consumption — without necessarily adding more hardware.
Can Inverter Data Reveal Solar Problems?
An Australian study has explored whether inverter data alone can spot solar production underperformance. Meanwhile, a separate study in Switzerland looked at estimating household electricity use from smart meter readings without dedicated CTs (current transformers). Together, the research hints at a future where software could give households a clearer view of what’s really happening in their systems, potentially helping them get better value from their solar.
Why Monitoring Matters
The “holy grail” for solar monitoring is seeing both solar generation and household consumption in real time, spotting system issues early, and getting tips to improve self-consumption. Most system owners only see solar production through their inverter app, which means performance issues and optimisation opportunities can easily go unnoticed. The stakes are even higher for households with batteries, where knowing when energy is generated, stored and used can directly affect savings.
Today, full visibility generally requires either CTs installed in the switchboard or a third-party monitoring platform that integrates smart meter data. Many users don’t see enough value to justify extra hardware or subscriptions, particularly if the system appears to be running smoothly. Standard inverter apps still mainly focus on solar PV output.
Aussie Research on Solar Production Monitoring
Scientists from the University of Technology Sydney, energy resources company Diagno Energy, and the University of New South Wales contributed to a new Australian study analysing AC-side solar production from over 1,000 residential systems across multiple regions. Using only standard inverter data, the researchers developed a rule-based workflow that compares expected daily output — based on system specs and local weather — with actual generation, flagging deviations that indicate common faults like inverter trips, extended zero output, or clipping.
This approach is novel because it can detect and classify underperformance without extra sensors or high-resolution measurements. The method identified most major faults with high accuracy — showing that, at least in this dataset, existing inverter data can be put to better use.
Overall, the study shows that manufacturers and monitoring platforms can use this workflow to spot problems and improve performance for systems already in the field. It’s a useful, incremental step toward smarter solar generation diagnostics, though it doesn’t provide appliance-level or real-time consumption insights.
Research on Consumption Data Without Extra CTs
Meanwhile, researchers at Switzerland’s Urban Energy Systems Laboratory explored whether ordinary smart meter data, recorded at 1–15 minute intervals, can estimate energy use for individual appliances — a technique known as non-intrusive load monitoring (NILM). They tested six algorithms — from statistical methods to modern deep-learning models — using different datasets and recording intervals.
The study found that deep-learning models like SGN and Seq2Point outperformed older methods, picking up both regular and changing patterns in appliance use more accurately. Performance still varied depending on appliance type and data frequency, showing that very low-frequency smart meter data can provide useful—but not perfect—appliance-level readings.
This research is relevant to Australia because it suggests that standard smart meters could one day give households a more detailed understanding of consumption without extra CTs or sensors, although the technology is still experimental.
How It Compares to Monitoring Right Now
In Australia, electricity networks already record household consumption through smart meters for billing and grid management, but these meters don’t provide real-time self-consumption or appliance-level visibility to the household.
Some monitoring platforms can combine inverter production with household consumption data, either via CTs or by accessing smart meter totals. These services still require additional setup and sit outside standard inverter apps.
Commercial platforms like Solar Analytics can show total household consumption alongside solar production, sometimes without additional hardware. However, they can’t break consumption down to individual appliances without CTs or other sensors, making the NILM approach in the Swiss study unique in ambition and scope.
The Takeaway
Combined, these studies point toward smarter, more data-driven solar monitoring. For now, getting the full picture still requires extra hardware or a dedicated platform, while standard inverter apps focus only on solar output.
Better software could one day give households a clearer sense of how their systems are actually performing. It may also give energy retailers a fresh opportunity to package “smart diagnostics” as a paid upgrade.
For those seeking deeper monitoring today, a dedicated platform remains the most practical option. To compare solutions, see SolarQuotes’ guide to solar monitoring.
About Kim Wainwright
RSS - Posts
Yes, it is hard to adjust, fine tune or adapt your power use if you can’t see what it actually is.
Even for people without solar / batteries, electricity retailers providing real time consumption data from the houses smart meter would go a long way.
Hi Andrew,
I don’t know if you remember the show Carbon Cops but when it was on they had a 5 segment light display on the wall of their demonstration houses. It showed real time electrcity consumption using traffic light colours depending on the load in the house, from green/good through amber to red/bad.
There was such demand we could have sold one with every solar power system we installed, but at the time nobody was selling it as a product sadly.
For most households, such monotring would need to be autonomous, alert type messaging for potential problems.
Most treat solar / batteries as set and forget, and that is risky for sure.
Without my SE app, I would not have seen (or known about) problems post install with incorrect wiring, a SAPN compatibility test fail for flexible exports, one inverter failure, another SAPN compatibility test fail, and in between all that, I had to have some optimisers reset remotely that had for some reason gone out of sync with the others.
All in all, it’s certainly not install solar and battery then forget them, expecting them to just work reliably.
If I had not watched this all closely, each day checking all ok and picking things up, no one would have, not SE, not the solar company.
I download SAPN data daily too, working on a spreadsheet to allow easy copy paste comparison to bill data / costs.
Eventually I will just do this quarterly for billing, and daily app checks to ensure all working ok.
Many solar estimation sites can provide a free tier for home owners. They will take your exact location and array orientation and will provide an estimate of what output you should see for any specific day. Its the case that their models that they expose are not the same as they provide to the local MW solar farm who pays them lots per year to get it exactly right, but over time if your actual output significantly underperforms against forecast then time to look at what’s wrong. Home Assistant has one out of the box, but there are alternates. I personally use Solcast. For their free home tier you can only have 1 array, for the out of box i think you can have as many as you need to reflect your actual array setup. While I use the solcast forecast with my HA to determine when and whether I turn on my storage water heater etc you can just log i to their portal and look at the forecast as a CSV file totally without HA if you don’t want it.
Opening another browser tab, to view my Victron system’s “Remore Console” over LAN, and double clicking on the DC-coupled solar icon, I see second-by-second yield for three related arrays in today’s milky total overcast:
3.9 kW W-facing 40° tilted: 685 W
6 kW N-facing 40° tilted: 944 W
6 kW S-facing 17° tilted: 242 W
The two AC-coupled strings are throttled down, f = 52.6 Hz.
As the equinox approaches, the S-facing array’s production sags, even when aided by cloud dispersion of light. It also displays voltage and current for each string in real time – full disclosure.
To mount a study to provide an obvious answer to a facile question seems comical. If you insist on detailed reporting, and can fund the decision, then it has always been available.
Mine tells me when an evening tree shadow impacts the W-facing array in summer – so long as production is not shut down due to negligible demand, and full battery.
But engineering data from consumer kit? Buy better, I suggest.
Seems to me (and I know bugger all) that if you ensure that you oversize the panels then you are getting a better % than you would have expected and solving most problems with usage.
The other thing is having a display that has a 5 segment light display on the wall of their demonstration houses. It can show real time electricity consumption using traffic light colours depending on the load in the house, from green/good through amber to red/bad. now make that compulsory just like circuit breakers in all new builds and people will become aware without effort saving money and the environment.
Steven,
You’ve hit the nail on the head. Installing triple the array size needed for summer maintains adequate yield in winter, when insolation falls to a third. Panels are still cheaper than batteries, and enough PV to drive the HWS in overcast demonstrates how production beats storage every time. (Mine is doing that now, with f = 52.2 Hz, so excess available. It’s the 8th consecutive overcast or cloudy day, but battery is at 100% SoC.)
I’ve seen various vendor’s system displays on this site – they show house load and grid flow, basic, but enough to understand usage, I figure. Victron’s CerboGX touchscreen shows instantaneous solar yield (AC & DC coupled), battery SoC, [ex|im]port, and individual string yield/voltage/current can be interrogated.
Products are available, matching the understanding and interest of different market segments. It is all there now, but installers can’t be across every product – especially configuration. (And budget is an issue.)
Oooh appliance-level monitoring, that’s a bit of a Holy Grail, have always wondered how they might crack that without flooding a home with clamps on every lead.
I think with existing software many can do a fairly good approximation of what’s using their electricity at a given moment. You get to see the patterns, ‘o that must be the stove’, or ‘has one of you kids left an AC unit on somewhere?’ Etc.
My fancy new washing machine does “AI” loads where it works out the wash cycle for what i have tossed in it and sends me messages when the washing is ready to hang up. I suspect passing on power usage data isn’t much of a stretch from that.
I had to run a load of washing this morning so i had a closer look at the app for the washing machine – blow me down – it does tell you how much power you use, as well as giving you stats for power use based on day, week and year/s!
At the start of Summer I got up on the roof to give the panels a clean. First time in 4 years for the majority of them. We live in a wet part of Sydney and the panels face north so while dust is an issue it doesn’t really stay for long without a good drenching each month. Lichen is more of an issue and spots had formed on many of the panels. After the clean (and the filthy black water was a testament to how dirty the surface was) the panels produced around 5% more through the day.
Given how long between cleans and how dirty the water was, it is surprising how good the panels are for collecting usable light.
As I have a microinverter system I can monitor the performance of each panel individually. Not only did the peak power of each panel improve, they started getting their first watts of generation a few minutes earlier and the same at the end of the day.