Cameras in Solar Module Plants: Cell Handling, Lamination and ALMM Audits
In a solar module plant, cameras do not test quality — electroluminescence does. Cameras catch the causes: the matrix lifted by one corner, the stack dropped at cell unloading, the laminator door opened early. EL tells you a module is cracked. Video tells you which station, which shift and which habit cracked it.
If you run a 300 MW to 1.5 GW module line in Gujarat, Tamil Nadu, Andhra Pradesh or Rajasthan, the pain of this business is specific: your capacity is new, your operators are newer, your buyer audits harder than any buyer you had in your last industry, and every point of yield you lose lands on a thin per-watt margin. This page is about what a camera layer honestly does and does not do on that floor.
Key points
- ALMM is now a two-list regime and the audit pressure is real. MNRE runs List-I for modules and a separate List-II for solar cells, with no blanket extension of List-II beyond 1 June 2026 and a limited window to 31 December 2026 for commissioning net-metering and open-access projects (MNRE ALMM page). Enlistment rests on a declared factory and a declared annual capacity — so your floor, not just your datasheet, is part of what is checked.
- Micro-cracks are a handling problem before they are a quality problem. Cell breakage and matrix mishandling cluster at a few transfers: cell unloading, stringer output, layup/glass loading, trolley movement and post-lamination handling into framing. All are camera-visible; none are EL-visible until the damage is already laminated in.
- Lamination discipline decides warranty exposure, not just scrap rate. A module that leaves an under-cured laminator will usually still pass flash test. It fails years later in the field. IEC 61215-1:2021 qualifies a design, not your Tuesday night shift (IEC 61215-1:2021, Ed. 2.0).
- The rupee case is arithmetic you can redo with your own numbers. At an illustrative ₹15/Wp realisation, a 580 Wp module is about ₹8,700. A 500 MW line is roughly 860,000 modules a year, so one percentage point of output is about 8,600 modules — around ₹7.5 crore of module value. Your actual loss is a fraction of that: a downgrade costs the price step between bins, a scrap costs the full value, rework costs labour and delay. Even at a quarter of the headline, a 30–40 camera AI layer is a low-single-digit-lakh monthly line item against it.
- Cameras cannot see the things that cause most of your yield loss. They cannot see cell quality, EVA gel content, string resistance, busbar solder strength, or the inside of a laminator. Any vendor telling you otherwise is selling an EL replacement that does not exist.
- Video is evidence, and evidence cuts both ways. Retained footage showing a documented handling failure can be produced against you in a warranty dispute. Decide your retention policy on purpose, not by NVR default.
Why ALMM made your shop floor an audited object
Before ALMM, an Indian module maker sold on price and a test report. Now enlistment ties a model to a named factory with a declared annual capacity, and MNRE has continued to revise the lists through 2026, including the separate cells list (MNRE ALMM page). That changes what an inspector walks in to look at. They are not only checking a lab result; they are checking that the line described on paper is the line running in the building, at the throughput claimed.
Two practical consequences for the owner:
First, process traceability now has commercial value. If you can show which line, shift and batch made a module, you can defend a claim instead of settling it.
Second, buyer and third-party audits become routine. Utility-scale developers and PSU buyers increasingly place third-party inspectors on your line during production. A floor that behaves the same whether or not an inspector is present is worth real money. A floor that only behaves when someone is watching is a liability that will eventually be filmed on someone else's phone.
Note what this section does not say: no ALMM order requires cameras. Cameras are how you make the claim about your own line checkable, not a compliance item.
The micro-crack chain: the transfers that quietly cost you
Micro-cracks are the classic silent killer because the module still works when it leaves your gate. The crack propagates under thermal and mechanical cycling in the field, and the power loss shows up years later, in someone else's O&M report, with your brand on it.
Here is where handling introduces them, and what a camera can actually detect:
| Station | Failure behaviour | Camera-detectable? | What a camera layer flags |
|---|---|---|---|
| Cell unloading / tray transfer | Cells fanned, dropped, mishandled; trays stacked too high | Yes | Drop events, stack height, ungloved handling |
| Stringer output → matrix transfer | Matrix lifted by hand at one edge instead of by carrier | Yes | Manual lift events at a station where lifts should be zero |
| Layup / glass loading | Glass slammed, matrix repositioned by hand, alignment forced | Partly | Rework touches per module, time-at-station spikes |
| Buffer / trolley movement | Trolleys pushed fast over a floor joint or drain channel | Yes | Speed and route violations in the aisle |
| Post-lamination → framing | Hot laminates stacked before cooling; corner impacts at framer infeed | Yes | Stack-before-cool events, impact/jerk at framer infeed |
| Cell interior, solder joints, EVA cure | Any of it | No | Nothing — this is EL, IV and gel-content territory |
The point of the table is the last row. A camera layer is a behaviour instrument, not a metrology instrument. Its job is to supply the causal half of the picture your EL station cannot produce.
The workflow that pays for itself is boring: EL flags a cluster of cracked cells in a batch. You pull the batch timestamp. You pull 90 seconds of video at each transfer for that window. In a plant with no video, that investigation takes a supervisor two days and ends in "operator error, counselled". With video, it takes ten minutes and ends with a specific fixture, a specific trolley wheel, or a specific 2 a.m. handover habit. See /efficiency/daily-shift-report-factory-owner-india/ for how that lands on your phone rather than in a file.
Lamination: cycle discipline is a camera problem
Laminator behaviour is where camera value gets misunderstood. You cannot see inside the chamber. What you can see, and what matters:
- Door opened before cycle completion. The most common shortcut on a line under throughput pressure at month end.
- Loading interval drift. If your laminator runs a fixed cycle and the loading interval starts shortening on night shift, someone has changed something.
- Laminate handling while hot. Stacking or edge-trimming a laminate before it has cooled adds stress to exactly the modules that already had a marginal cure.
- Trim and edge-clean quality. Visible, and worth correlating with later delamination complaints.
None of this replaces gel-content sampling or peel testing. It tells you whether your sampled result represents the whole shift or only the hour the QC engineer was standing there. IEC 61215-1:2021 is a design qualification and type-approval standard — it says nothing about whether your line held discipline last night (IEC 61215-1:2021, Ed. 2.0).
What cameras honestly cannot do here
This is the section vendor blogs skip.
Cameras cannot grade cells. Incoming cell quality — especially with imported cells or newly commissioned domestic cell lines under List-II — is a large single yield variable, and video contributes nothing to it.
Cameras cannot see micro-cracks. A hairline crack in a laminated cell is invisible to a visible-light camera at any realistic mounting distance. Anyone demonstrating "AI crack detection" from a ceiling camera is showing you EL images relabelled.
Cameras cannot measure cure. There is no practical optical proxy for cross-linking on a production floor.
Cameras will not fix a fixture problem. If your stringer output carrier is the wrong geometry, video only tells you faster that you have a mechanical problem you must spend money to correct.
And cameras can hurt you. Two specific ways. One: footage becomes evidence. If a buyer alleges systemic handling failure, retained video is a record of your own knowledge — decide retention deliberately (/compliance/cctv-footage-retention-period-india-factories/). Two: a badly introduced camera layer reads as surveillance to a skilled workforce that is scarce in the Sanand, Mundra and Sri City belts. Announce it as a scrap-and-yield tool, publish a worker notice, keep cameras out of changing rooms and rest areas, and handle personal data properly under India's DPDP framework (/compliance/worker-cctv-privacy-notice-dpdp-india/).
On safety law, the Factories Act 1948 was repealed on 21 November 2025 when the OSH Code 2020 came into force, with Central Rules notified on 8 May 2026 — and your operative rulebook is your state's OSH rules once notified. Hazardous-process and welfare duties apply to your lamination and chemical-handling areas either way. Cameras support those duties; they do not discharge them (/compliance/osh-code-2020-factory-cctv-india/).
A workable camera spec for a module line
For a single 500 MW line, a sane starting layout is roughly 28–40 cameras:
- 6–8 at cell unloading and stringer output — the crack-origin zone, mounted close, at 4–6 MP, looking down at the transfer plane, not across the hall.
- 4–6 at layup and glass loading.
- 4 at the laminator bank, including one framing the doors and one on the outfeed cooling area.
- 4–6 at framing, junction-box and curing.
- 4 at EL/flash test and the rework/downgrade bench — the highest-value cameras in the plant, because this is where downgrades get argued about.
- 4–6 at inbound cell stores, outbound module stores and the weighbridge or dispatch dock.
Mounting matters more than megapixels here: cell-handling stations need close, high-angle views, which usually means machine- or gantry-mounted units rather than corner domes. Budget PoE properly (/cameras/poe-budget-power-factory-cameras-india/) and decide the recording architecture before you buy (/cost/nvr-vs-cloud-vms-factory-video-storage-india/).
For what a full AI layer costs in rupees, capex versus monthly, see /cost/factory-camera-ai-system-cost-india/. The honest test of whether it is worth it: if you cannot currently trace an EL-flagged batch back to a station within an hour, the system pays. If you can, it probably does not.
What to do in your first month
- Instrument the crack-origin transfers only. Do not wire the whole plant.
- Timestamp-sync cameras with your MES or batch log. Without this, video is decoration.
- Run one month of EL-to-video investigations and record what you find. This is your business case, in your own numbers.
- Publish the worker notice and retention policy before, not after, go-live.
- Only then widen to stores, dispatch and safety zones.
FAQ
Does ALMM require CCTV in a solar module plant? No. ALMM enlistment concerns models, manufacturers, declared factory capacity and inspection, not video systems (MNRE). Cameras help you survive an audit by making your process traceable and consistent, but no ALMM order obliges you to install them. Treat any vendor claiming an ALMM camera mandate as unreliable.
Can cameras detect micro-cracks in cells? No. Micro-cracks in encapsulated cells are detected by electroluminescence imaging under forward bias, not by visible-light cameras. What cameras do is identify the handling events that create cracks, so your EL rejects fall over time instead of merely being counted more accurately.
Will cameras replace my EL and flash tester? No. EL and IV measurement are metrology; cameras are behaviour observation. The value comes from joining the two: EL says which batch is bad, video says which station and shift made it that way. Any proposal positioning cameras as a test-equipment substitute should be rejected outright.
What about worker privacy on a shift-heavy module line? Publish a written notice, restrict cameras to process and material areas, avoid rest, changing and medical spaces, and set retention to the shortest period that serves your quality and dispute needs. See /compliance/cctv-banned-areas-factory-india/. Doing this properly also helps if a labour inspector or a buyer's social-audit team asks.
We are commissioning a new cell line under ALMM List-II. Different camera needs? Yes — cell lines are cleaner, more automated and more sensitive to particulate and handling at wafer load, texturing transfer and cassette movement. Camera count is lower and placement is tighter to the machines. On List-II timing, MNRE has stated no blanket extension beyond 1 June 2026, with a limited window to 31 December 2026 for commissioning net-metering and open-access projects (MNRE) — check the current order before you plan around it.
How many cameras does a 1 GW module plant need in total? Typically 60–90 including stores, dispatch and perimeter, but the number is the wrong question. Two well-placed cameras at stringer output beat twenty corner domes. Start from the failure you want to see, not from floor area — the method is in /cameras-and-placement/how-many-cameras-factory-floor/.
