AI CCTV for solar farms and battery storage sites

Cable theft from solar and battery storage sites is rising fast

Renewable energy sites hold large amounts of copper cable, spread across remote fields that nobody watches at night. Crime intelligence shared with UK police forces shows organised groups targeting them more often, and coming back to the same sites. Every figure below links to its source.

DeterTech, which has a data sharing agreement with all 43 police forces in England and Wales, reported a 59 per cent year-on-year increase in offences against solar, onshore wind and battery storage facilities, with 169 offences in 2025. In Energy Global it describes the threat as a planned, organised one rather than opportunistic: groups working across several sites in one campaign, often in clusters a short distance apart, and targeting grid connection points, cabling routes and access infrastructure. The value of copper has risen by around 25 per cent since late 2025, which makes cable an even more attractive target.

Individual thefts can be large. Wiltshire Police warned that around 30km of cabling, worth around a hundred thousand pounds, was stolen from a solar farm near Trowbridge, and said they believed thefts like it were linked to organised crime groups (Wiltshire Times). The cost does not end with the cable: strings and inverters taken offline stop generating until the damage is repaired.

Repeat targeting and decoy tactics

Sites that have been hit once are likely to be hit again. In a 2025 media alert, DeterTech warned that vulnerable sites that do not get an immediate and visible security upgrade are highly likely to be targeted again, and described one location suffering three break-ins in a month. The same alert described thieves in Derbyshire luring mobile patrols to a nearby location in a co-ordinated distraction, then stealing string cable from the site the patrols had been guarding.

The way the cable is taken matters too. A DeterTech analyst told Solar Power Portal that string cable is the most common target, and that thieves pull the strings out row by row. With a hundred or more panels in a row, a gang can collect a large amount of cable quickly, and every row it strips stops generating.

Organised groups are targeting grid connection points, cabling routes and access infrastructure, where disruption is both easier and more valuable.Summary of DeterTech crime intelligence, Energy Global, June 2026

All of this points to the same gap. A solar farm is large, remote and dark, so a gang can be on site for a long time before anyone knows. Patrols can be drawn away, and a site that relies on someone noticing the damage the next morning has already lost the cable. The cameras a site already has can close that gap, if something is watching them at two in the morning.

Where a solar farm is most exposed, area by area

Select an area on the plan, or use the buttons, to see the risk there, what a camera needs to see and where the detection zone goes. The dashed outlines show where zones would be drawn.

Polygon zones drawn along the fence line

Most solar farm boundaries run for long distances beside fields, hedges and lanes. A camera that looks along the fence sees a long stretch of it, but it also sees the field next door, the road and the trees. The detection zone has to follow the fence itself.

In Woodcock AI you draw polygon detection zones on each camera view, point by point. On a fence line, the zone runs along the tops of the fence posts and down to the base of the fence, and stops at the boundary. A person climbing the fence or a section being lifted is inside the zone; a hare in the next field, a car on the lane or a tractor working the neighbouring farm is not. A simple box drawn over the view cannot follow a fence that runs diagonally away from the camera, which is why polygons matter so much on long boundaries.

Each zone can have its own schedule, so the fence line is armed once the O&M team has left and stays armed until the morning. During the day, when contractors and landowners are about, the same camera can stay quiet or watch only the most sensitive stretch. This is intrusion detection fitted to the shape of the site rather than to the shape of the camera view.

Vehicle detection on access tracks and field gates

Cable is heavy and bulky, so a theft almost always involves a vehicle. Vans, pickups and trailers use the same access tracks and gates as the maintenance team, which makes those tracks one of the best places to catch a theft early.

A camera on the main gate and the first stretch of the access track can be set to look for vehicles out of hours. Vehicle detection sends an alert as soon as a van or pickup appears on the track when nobody should be there, with the clip, so the person on call can see whether it is a contractor who arrived early or a vehicle with no reason to be on site. A line-crossing rule across the entrance adds a second check: a vehicle or person crossing the line into the site raises an alert, whichever way they came.

Vehicles parked on a lane outside the fence matter too. Gangs often case a site before a theft, or leave a vehicle close to the boundary while they work. A zone on the verge beside the fence, armed out of hours, can flag a vehicle that stops there in the middle of the night. Our metal theft prevention guide covers the wider pattern of cable and metal theft and how AI CCTV helps.

Inverter, transformer and substation compounds

Inverter and transformer stations, and the substation at the point of connection, are where the most valuable equipment and cable on the site come together. Damage there takes whole sections of the site off line.

Most compounds sit inside their own palisade fence with a locked gate. A camera on the compound can have a zone across the gate and a second along the fence, so anyone opening the gate or climbing the fence out of hours raises an alert. Because zones are drawn as polygons, the zone on the gate can follow the gate itself, without taking in the floodlight, the kiosk or the panel rows behind it.

The same camera can watch for smoke and flame at the same time. Inverters and transformers are electrical equipment, and a fire in a compound is a serious event at any hour. Smoke and flame detection can run around the clock on that camera, while the intrusion zone on the gate runs only out of hours.

Early smoke and flame detection for battery storage containers

Battery energy storage systems add a second risk to a renewable site. A fault inside a lithium-ion battery can lead to overheating, smoke and fire, and a battery storage compound is often unstaffed and remote.

Woodcock AI watches the cameras that cover battery containers for the first visible smoke and flame, and sends an alert with the clip to the people you choose. The clip matters. The person on call can see straight away whether it is smoke from a container roof, mist drifting across the compound or dust kicked up by a vehicle, and decide whether to call the fire service, alert the battery operator or send someone to look. Every event is saved with its clip for the investigation that follows.

Camera detection is a complement to the fire safety built into the site, not a replacement. Battery containers have their own monitoring, detection and suppression, and the site will have its own fire strategy agreed with the fire and rescue service. Video AI adds a view from outside: it covers the whole compound, the gaps between containers and the ground around them, and it can see smoke that drifts away from a container before it reaches a sensor. We cover the wider risk of lithium-ion fires on our lithium-ion battery fires page.

For the best view, cameras should see every container face and the roofs, with working lighting or infrared at night. The free site assessment checks which of your existing views already do this and where a camera would need to move or be added. For more on how smoke is detected on ordinary CCTV, see our video smoke detection guide, and for wind, rain and night views outdoors, our outdoor fire detection guide.

Self-learning AI that knows sheep, deer and hedges are normal

A solar farm is a hard place for a generic detection model. Sheep graze between the rows, deer and foxes cross at night, hedges move in the wind and headlights from a nearby road sweep across the fence. A model that alerts on all of that is soon ignored.

Woodcock AI is built for each site. Each camera gets a model fitted to its own scene, trained on hours of that camera's footage where nothing happened, so it learns what normal looks like there. Polygon zones keep detection on the areas that matter, and schedules keep the intrusion zones quiet when the O&M team is on site.

Every detection has a feedback button. If an alert was not useful, mark it and the feedback goes into retraining for that camera. Over time the AI gets better at telling a grazing flock from a person in the rows, or a deer at the fence from someone lifting it. Read more about self-learning AI, and for practical fixes to lenses, lighting and night views, see how to reduce CCTV false alarms.

Ask your cameras what happened, in plain English

After a theft, the first questions are always the same. When did they arrive, how did they get in, and what vehicle were they using? Scrolling through a night of footage from a dozen cameras to answer them takes hours.

With Woodcock AI you can ask your cameras questions in everyday language, and get back the clips that answer them. Every alert also comes with a short plain-English description of what the camera saw, so the record of the night reads like a log rather than a pile of recordings. Download the clips and hand them to the investigating officer, or to your insurer.

Good evidence depends on good views. A camera that sees a vehicle's arrival at the gate, and the track it took, is far more useful than one that only sees the panel rows. The National Business Crime Centre's infrastructure crime responders guide notes that fingerprints and DNA on cut cable can help identify those responsible, which is one more reason to keep a scene untouched until the police have seen it.

Who hears about it, and how

An alert is only useful if it reaches someone who can act. Woodcock AI sends each alert to the people you choose for that camera and that time of day, with the clip and a plain-English description.

Alerts go out by the Woodcock AI app, as a push notification with the clip, and by email. Alerts to a monitoring centre (ARC) and phone-call alerts are available as optional add-ons. For a remote solar farm, the choice often depends on who is responsible out of hours: an asset manager, an O&M contractor, a security company or a monitoring centre. You can send fence and track alerts to one group and battery compound alerts to another, and change the rota as the season changes.

One layer in a site's security, not the whole of it

No single measure stops cable theft. The sites that fare best combine physical security, deterrence, detection and a clear response. Video AI on existing cameras is the detection layer that works across the whole site at night.

• Physical securityFencing, gates, locks and anti-climb measures that slow intruders down and keep vehicles off the site.
• Forensic marking and signageMarking cable and equipment so stolen material can be traced, with signs telling would-be thieves it is marked.
• Video AI on existing CCTVWoodcock AI watches fence lines, tracks, compounds and battery containers, and sends the clip to the people who can act.
• Battery system fire safetyThe detection, suppression and shutdown built into the battery containers and set out in the site’s fire strategy.
• People and proceduresGate checks, an out-of-hours call-out plan, and reporting every incident and suspicious vehicle to the police.

How to add AI to your site's existing cameras

Most solar farms and battery storage sites already have CCTV on the entrance, the compounds and parts of the fence, often installed when the site was built. Adding AI to those cameras is usually a matter of connecting, zoning and choosing who is alerted.

Woodcock AI works on sites with a network connection to their cameras. How that connection is made differs from site to site, so the free site assessment looks at your recorder, your cameras and your connection before anything is set up, and tells you plainly what will and will not work.

1. 1

   Free site assessment

   We review your camera views, recorder and network connection, and agree which areas matter most: fence lines, gates, tracks, compounds and battery containers.

2. 2

   Connect your cameras

   Woodcock AI connects to your existing IP cameras or recorder over ONVIF or RTSP, through an on-site edge server or the cloud.

3. 3

   Draw the zones

   Polygon zones are drawn on each camera, along a fence, across a gate or over the containers, with line-crossing rules where a track meets the boundary.

4. 4

   Set the schedules

   Each camera gets its own schedule. Intrusion zones can run out of hours, while smoke and flame on the battery compound are watched around the clock.

5. 5

   Choose who is alerted

   Pick who gets each alert by app and email, with the clip. Add a monitoring centre or phone-call alerts if nobody is on call.

6. 6

   Let it learn your site

   The AI is fitted to each camera, and the feedback button on every detection goes into retraining, so it keeps learning what is normal on your site.

Need more cameras, or a new system for a new site? We offer design, supply and install, from site survey and design through supply, installation and commissioning to go-live and ongoing support. Read how it works for the technical detail.

For installers, security companies and O&M contractors

Many solar farms and battery storage sites are looked after by an O&M contractor or a security installer who already maintains the cameras. Those companies can add Woodcock AI to the systems they look after.

Through the installer programme, partners connect their customers' cameras, draw the zones and set the schedules, then manage every site from one place. It turns cameras that only record into cameras that raise the alarm, without replacing the hardware. Our guide to adding AI to existing CCTV for installers explains how it works in practice, and the same approach applies on industrial sites and construction sites.

Reduce your site's cable-theft exposure: a ten-point check

Tick what is already true on your site. Your answers are saved in this browser only. Several points follow the National Business Crime Centre's infrastructure crime responders guide, which advises investigating unlocked gates and out-of-hours work, and always reporting suspicious activity and crimes to the police.

• Every gate and access track has a camera that sees vehicles arriving
• Fence lines out of sight of the road are covered, not just the entrance
• Inverter, transformer and substation compounds each have a view
• Battery containers can be seen on every face, day and night
• Night views have working infrared or lighting
• Gates and locks are checked, and an unlocked gate is always investigated
• Someone is named to receive alerts out of hours
• A visible security upgrade follows any incident
• Cable and equipment are forensically marked, with signs saying so
• Every incident and suspicious vehicle is reported to the police

If you find cut cable, the NBCC guide says secure the scene and do not touch it: it may still be live, and fingerprints and DNA on it can help identify those responsible. Anything unticked above is worth raising in a free site assessment. For wider out-of-hours cover, see the out-of-hours site security checklist.

Sources

We only quote figures we can trace to a public source. Published 26 October 2026 by Woodcock AI.

1. Securing the solar surge Energy Global, 17 June 2026, with DeterTech crime intelligence. Last checked 26 October 2026.
2. Media alert: organised criminal gangs targeting UK solar farms in surge of spring thefts DeterTech. Last checked 26 October 2026.
3. UK entering peak period of solar PV theft by fully industrialised criminal network Solar Power Portal. Last checked 26 October 2026.
4. Trowbridge solar farm sees 30km of cables stolen Wiltshire Times, reporting Wiltshire Police. Last checked 26 October 2026.
5. Infrastructure crime: responders guide National Business Crime Centre (NBCC). Last checked 26 October 2026.