Closed Poultry House Monitoring: Alarms and Data

A closed poultry house, where fans alone keep tens of thousands of broilers alive, needs monitoring built on one contract: detect within minutes a stopped fan, a power cut, or a generator that never started; wake a person who acknowledges the alarm; and keep the climate automation running on site when the internet drops.

Graphs, harvest reports, and feed-efficiency records help assess the system's value over time. Operational protection comes first: a night-time ventilation fault needs to be detected and handled even when nobody is watching the dashboard.

Why does a closed house turn minutes into money?

In a closed house (Indonesian farmers call it a close house), the walls are sealed. Air moves only because exhaust fans pull it in through wet cooling pads at one end and push it out at the other. The building is on life support. Cut the power and nothing replaces the fans, so the body heat of thousands of birds stays inside.

Auburn University's poultry engineers put it plainly in their July 2002 extension newsletter: if utility power fails in hot weather and no standby generator takes over, "birds will begin dying within minutes."

Indonesia watched that sentence play out twice in one week this year. On 24 July 2026, a six-hour rolling blackout hit a broiler farm in Hulu Sungai Tengah, South Kalimantan, and the farm's generator failed at the same time. Kompas reported that 5,000 of its 8,000 birds died, all of them at market weight. The loss was estimated at Rp250 million, or Rp50,000 per bird.

Five days later, farmers in Palangka Raya dumped 1,500 dead chickens at the local PLN office after repeated outages. The head of the Central Kalimantan poultry association told Tabengan that about 90% of chicken houses in the province are closed houses, and that a 40,000-bird farm can lose up to Rp360 million in a single mass-mortality event.

The Hulu Sungai Tengah case shows that having a generator does not ensure it will work when needed. Monitoring and regular tests need to establish whether the backup is ready.

What should the sensors actually watch?

Start from the failures that kill birds, not from a sensor catalogue. The Auburn newsletter lists the usual causes when a house suddenly overheats: a broken fan belt, a burned-out fan motor, a tripped fan breaker, a failed thermostat, a stuck air inlet. Every row below exists to catch one of those, or the power problem behind them.

Two rows need more explanation.

Measure fans by current, not by command. The controller knows it switched fan group 6 on. It does not know the belt snapped at midnight. Current sensors flag load changes, but current alone does not prove blade rotation or an intact belt. For critical fans, combine with tested rotation, airflow, or pressure sensing. The same logic applies to commands sent from a phone: the app should show two separate confirmations, one that the command arrived and one that the equipment actually changed state.

Treat gas sensors as trends, not verdicts. As a reference point, the EU broiler welfare Directive 2007/43/EC caps ammonia at 20 ppm and CO2 at 3,000 ppm, measured at the level of the chickens' heads, for covered EU farms stocked above 33 kg/m². This is a reference for the design discussion, not an Indonesian legal requirement. Affordable ammonia sensors drift in a humid, dusty house. Calibrate them during the empty break between flocks, and alarm on a sustained trend rather than a single reading.

Where should the automation run?

On site, in the control panel. Never in the cloud.

Internet access on a rural farm usually means a 4G signal that comes and goes. If the logic that turns fans on during a heat spike lives on a server in Jakarta, every signal dropout becomes a ventilation failure. The rule is simple: the local panel makes every automation decision, and the internet only carries readings out and remote commands in. When the signal disappears, the fans keep following the last rules they received.

Farms reach this architecture from one of two starting points.

The house already has a climate controller. Keep it. Add the gateway, a small box that reads the controller and relays its data to the internet. It connects either over Modbus, a common industrial protocol where the controller answers requests for numbered values, or through dry contacts, plain on/off signals taken from the relays (electrically operated switches) inside the panel. If the existing controller already manages the climate well, preserve that function; survey the connection before deciding whether wiring changes are needed. The site survey decides which connection the controller supports.

The house is still switched by hand. Then the panel, the sensors, and the software go in as one package, and the house becomes automated and monitored at the same time.

Either way, the gateway must keep readings while the link is down and upload them when it returns. Otherwise the history of the worst night in the cycle is exactly the part that goes missing. If the gateway buffers to an SD card, give it a write budget, because SD cards in field devices fail for predictable reasons. Everything from the gateway to the server is standard IoT plumbing, covered in the IoT backend guide.

A local dashboard served over the farm's WiFi completes the design. With the internet gone, the night guard can still open a phone and see every temperature and every fan.

How does an alarm reach someone asleep at 2 a.m.?

The most useful line in the Auburn newsletter is printed in capital letters: alarms do nothing to save your birds. They only bring a person to the house. So design the alarm path around that person.

The power-loss alarm must work without grid power. Use a battery-backed relay that sounds a siren and sends a notification when grid voltage disappears. It should fire whether the generator starts or not. You need to know the grid is down even when the generator covers it, because the next outage may find that generator out of diesel.

The local siren must not need the internet. Mount it at the guard post. A phone notification is the second path, not the first.

Every serious alarm needs an owner. If nobody acknowledges it within a few minutes, it escalates to the next number on the list, then the next. A realistic night: a fan group stops at 02:13, the operator on shift does not respond, the alarm escalates to the farm supervisor, and the breaker is reset at 02:21. The owner reads the timeline over breakfast.

Not every fault should wake someone. Auburn separates loud alarms from soft ones. If one of six temperature sensors fails while the other five keep control, that is a soft alarm: a note on the dashboard for the morning. Waking people for soft faults teaches them to ignore loud ones.

Silence is a fault. A gateway that stops reporting must trigger an alarm on the server within minutes. Otherwise a dead gateway looks exactly like a quiet, healthy night.

An optional SMS module adds one more path over plain cellular service, for the nights when mobile data is unreliable but calls and texts still get through.

Why must alarm limits change as the birds grow?

A day-old chick needs a warm house. A five-week-old broiler needs a cool one. The target temperature moves almost every day of the cycle, so a fixed alarm limit is wrong for most of it.

Auburn names the consequence directly: the most common cause of bird loss is human error, meaning alarms and backups that nobody maintained or reset. Its rule of thumb puts alarm limits 8–10 °F (about 4.5–5.5 °C) above and below the target temperature, checked daily. On a farm with six houses and two shifts, "checked daily" is exactly the step that gets skipped.

Software can calculate the daily limits, while staff still check that the settings and sensors are appropriate. Store the target temperature curve for each flock, record the day the chicks were placed, and let the system compute today's alarm band on its own. When someone overrides a limit by hand, log who did it and give the override an expiry time.

Backups need the same discipline. Run a drill during the empty break between flocks: cut grid power on purpose and confirm, in order, that the relay trips, the siren sounds, the generator starts and takes the load within seconds, the fans come back one group at a time so the generator is not hit by every motor at once, the notification arrives, and escalation reaches the second person. Record the result. The EU directive treats this as paperwork worth keeping: farms stocked above 33 kg/m² must document their alarm and backup systems and record technical inspections of ventilation and alarms.

Between flocks, the weekly generator test should appear in the data as a running signal, a voltage, and fan current. A generator test that leaves no record might as well be a rumour.

Using data to assess feed costs

Once the operational safety requirements are addressed, the same data can help assess feed costs. Margins are thin. In June 2026, the smallholder farmers' association Permindo told Kompas that live broilers in Greater Jakarta and Banten sold for Rp15,000–16,000/kg against production costs of Rp21,000–22,000/kg. Feed had risen Rp800/kg during 2026, to Rp8,800–9,400/kg.

At those prices, money hides in the feed conversion ratio (FCR), the kilograms of feed a bird eats to gain one kilogram of live weight. Improve FCR by 0.1 on a 2 kg bird and each bird eats 0.2 kg less feed, about Rp1,800 at Rp9,000/kg. Across a 30,000-bird house, that is Rp54 million per cycle.

Operators enter feed deliveries, sample weights, and daily deaths from their phones. The system does the arithmetic:

• FCR and average body weight, per house, updated as the cycle runs.
• Performance index (IP), the single score Indonesian integrators use to rank a flock: livability (the % of birds alive at harvest) × average weight in kg, divided by FCR × age in days, times 100. A flock with 95% livability, 2.0 kg average weight, and FCR 1.55 at 32 days scores (95 × 2.0) ÷ (1.55 × 32) × 100 ≈ 383.
• Hours outside the target temperature band, per house per cycle. This helps investigate poor FCR alongside feed quality, disease, and other factors.

Comparisons turn those numbers into decisions: why house B burned more feed than house A, which settings produced the best flock, and whether a hot week lines up with a jump in deaths.

Daily mortality is also an early warning in its own right. A 2024 study in Pathogens on avian influenza outbreaks in Dutch poultry cites a proposed reporting threshold for broilers: daily mortality above 0.17% after the first week. In a 30,000-bird house, that means more than 51 birds in one day. A system that knows the bird count can flag the line the moment the operator enters the day's deaths, not at the weekly recap.

Water works the same way, with a caveat. Birds that are getting sick often drink less before they look sick, which is why monitoring vendors watch water intake closely. A fixed threshold copied from another farm rarely fits. Compare each house with its own previous cycles at the same age, and treat a drop as a reason to walk the house, not as a diagnosis.

The data belongs to the farm. Put that in the contract, along with the export format.

What to ask before buying a monitoring system

1. If the internet drops for six hours, what keeps running and what stops?
2. Does the power-loss alarm still sound when the grid is down and the generator has also failed?
3. Is fan status measured from current, or assumed from the command?
4. What happens when the gateway itself goes silent, and how fast?
5. Who receives the alarm after the first person ignores it, and after how many minutes?
6. Do alarm limits follow the flock's age automatically?
7. How is a backup drill run and recorded between flocks?
8. Can the existing climate controller stay, and how will the gateway connect to it?
9. Who owns the data, and in what format can it be exported?

Ask for test evidence behind the answers to questions 1, 2, and 4. A dashboard demonstration alone does not establish whether the alarms and backup paths work.

The takeaway

Closed houses made broiler farming more productive by making it more dependent. Fans, grid power, and the generator are now the flock's lungs, and an unnoticed failure at night can cause a major loss.

Buy in the right order. First the alarm that works without grid power or internet. Then automation that lives on site. Then the data that helps assess costs and performance each cycle. The Hulu Sungai Tengah case shows that an available generator can still fail when needed. The reporting does not establish its full maintenance history. The design lesson is to test the backup under load and record the results before an emergency.