Find every fault in your solar fleet. Fix the ones that cost you.
Solvane AI reads drone imagery, checks it against string-level telemetry and puts a dollar figure on every defect. Your O&M team fixes what matters first, backed by evidence instead of a 200-page PDF.
This is the working MVP on a sample plant. Click around, run an analysis, or open it full screen.
Solvane AI is an inspection-intelligence platform for solar plants. Fly the site the way you already do, and Solvane works out which modules are faulty, checks each fault against what the inverters are reporting, and tells you what it is costing.
It examines every image before it says a word. Hotspots, failed bypass diodes, dead strings, cracks, soiling and vegetation: each finding is pinned to a block, row and string, given a confidence score and an estimated loss, and stored as evidence.
Asset owners, O&M contractors and performance engineers stop stitching together thermography reports, SCADA exports and spreadsheets. They work from one ranked list of what is wrong, what it costs and who is fixing it.
Watch one thermal flight turn into a repair decision.
The same models as the app, on one synthetic thermal capture of Block B1 at Mesa Verde Solar. Click a step, or let it play.
Solar plants generate data everywhere. The answers stay scattered.
Each season brings thousands of thermal and RGB images, inverter and string telemetry, work-order history and warranty files. Teams still piece them together by hand to answer one question: which faults are costing us money, and is anyone fixing them?
Sample season at Mesa Verde Solar (48 MWdc): 18,400 thermal frames · 3,312 strings · 6 blocks · 4 strings flagged by the anomaly model
Inspect. Understand. Compare. Prioritize. Maintain.
One pipeline turns a drone flight into a ranked repair list, and every step is something you can click through in the live demo.
Inspect
Upload RGB and thermal captures from any drone. Each batch is tagged to a site, block and inverter.
Understand
Vision AI segments every module and proposes findings, each with a location, evidence and a confidence score. Engineers confirm or dismiss.
Compare
Findings are matched against the last inspection and checked against string telemetry, so new faults and confirmed faults stand out.
Prioritize
Each finding gets kWh and dollars lost per year from your yield and PPA price, plus a 0–100 risk score and a P1–P4 priority.
Maintain
Confirmed findings become work orders with SLAs, exported to your CMMS or a signed inspection report.
One flight. Every answer about the plant.
Every demo below runs live in your browser on the same sample plant. Try them.
Your thermal flights become measurable change.
Drag across April's and October's thermal capture of Block B1. Solvane found a new dead string and a new bypass-diode failure, and confirmed that the hotspot from April is still there.
- Every module segmented and measured, not just the obvious ones
- Thermal treated as relative: no made-up temperatures
- Same-block comparisons across seasons
Drag the handle, or use the arrow keys.
Catch the faults the drone flew past.
An unsupervised model compares every string with its neighbours each day, after normalising for weather. Strings that drift away from the block are flagged, even between flights, and linked to the imagery when the two agree.
- Robust median/MAD z-score: one bad day doesn't trigger it
- Weather cancels out, because every string sees the same sky
- Every flag explains which string, how far off and since when
Know which faults are worth a truck roll.
A cracked cell and a dead string are both "defects", but they are not worth the same. Solvane converts each one into kWh and dollars lost per year using your specific yield, PPA price and module rating, then ranks the whole plant.
- Loss is estimated per finding, then totalled per block and site
- Severity, confidence and safety feed a 0–100 risk score
- The score maps to P1–P4 response SLAs your team can override
Ask your plant.
Copilot answers from this plant's findings, telemetry and inspection history, and cites the evidence. When the data can't answer a question, it says so.
- Numbers come from the record, never from the model
- Every finding ID it cites links back to the image
- Works offline, with a language model as an upgrade
See risk where it is, not in a spreadsheet.
Every block is coloured by its worst open finding and string anomaly. Click a block to see what is wrong and what it costs.
- Imagery findings and telemetry anomalies on one map
- Drill down from site to block, string and module
Five passes behind every upload.
Solvane is more than a chatbot over thermography PDFs. It measures every module, places each fault on the plant, checks it against telemetry and prices it, using rules an engineer can read.
Segments the image
Every module in the frame is isolated and measured: intensity, colour and texture, then the shape of any anomaly.
Finds the outliers
Robust statistics compare each module with the rest of the array, so a fault stands out against its own neighbours.
Places it on the plant
Each finding is linked to a site, block, inverter and string, alongside earlier inspections.
Cross-checks telemetry
String anomaly detection confirms or questions what the image shows, and finds faults the image missed.
Prices and explains
The loss model turns each finding into dollars a year, and Copilot answers cite the evidence behind them.
Seven modules. One asset record.
Solvane Vision
Detect hotspots, diode failures, dead strings, cracks, soiling and vegetation in RGB and thermal imagery.
See it working ›Solvane Strings
Weather-normalised string anomaly detection from inverter telemetry, linked to the imagery.
See it working ›Solvane Value
kWh and dollars lost per year for every finding, with risk scores and P1–P4 priorities.
See it working ›Solvane Copilot
Ask questions of the whole plant and get answers that cite the evidence.
See it working ›Solvane Work Orders
Turn confirmed findings into SLA-tracked work orders, CMMS exports and inspection reports.
See it working ›Solvane Fleet
Compare revenue at risk and response times across every site in the portfolio.
Learn more ›Solvane Forecast
Use inspection history to flag where degradation is accelerating before it shows in yield.
Learn more ›See all five Pilot modules working
Every Pilot module runs on the sample plant today. Preview modules ship to pilot partners next.
Open the live demo ›Built for everyone who answers for plant performance.
Which faults are costing us money right now?
Revenue at risk ranked across every site, with photo evidence behind each number.
What should the crew fix this week?
A prioritised work list with SLAs, and evidence that each repair was made.
How do we deliver reports in hours, not weeks?
AI-assisted, reviewed findings and white-label reports from the same upload.
Is this string actually broken, or just shaded?
Imagery and telemetry side by side, with the statistics behind every flag.
Is this plant being looked after?
A traceable defect history and response times for due diligence and claims.
Which sites need intervention?
The same measures across the fleet, from portfolio down to a single module.
Move from inspection reports to inspection intelligence.
Your plant data stays protected.
Owners trust Solvane with imagery and performance data that affect revenue and valuations. The platform is built for that.
Start with one site.
We are running early pilots with a small group of asset owners, O&M contractors and drone service providers.
Talk
A 30-minute call about your fleet, your inspection cadence and how faults get fixed today.
See it on your site
A live demo on the sample plant, then on one of your own flights and SCADA exports.
Pilot
One site for 60 days, with agreed success measures and weekly check-ins. $4,900, credited in full to your first year.
How does Solvane find defects in drone imagery?
It segments every module in the image, measures intensity, colour and texture per module, and flags modules that are statistical outliers against the rest of the array. A dead string shows up as a run of uniformly warm modules; a failed bypass diode as one warm third; a hotspot as a sharp local peak. Each finding is a proposal until an engineer confirms it.
Do we need special drones or radiometric cameras?
No. Solvane works with standard RGB and rendered thermal captures from any drone, and treats thermal as relative. Radiometric files can be added for calibrated temperature analysis on Enterprise plans.
How are the dollar figures calculated?
For each finding: modules affected × module rating × your site's specific yield × a typical loss factor for that defect type × your PPA price. They are planning estimates, labelled as such, and can be calibrated against metered data.
How accurate is it, and who has the final say?
On our synthetic benchmark the on-device engine found every planted defect with no false positives; real-world accuracy depends on flight quality and is measured during each pilot. A person always has the final say: nothing becomes a work order until it is confirmed.
Does it replace our SCADA or CMMS?
No. Solvane reads string telemetry from your SCADA and sends work orders to your CMMS (Maximo, SAP PM, Fiix, or CSV). It is the layer that connects them.
What happens to our data?
It stays yours. Each organisation's data is isolated, encrypted, and never used to train shared models without written opt-in. Core image analysis can run entirely on-device.
Know what's broken, and what it costs.
See Solvane on the sample plant, then on your own flight.