What Size Generator Does a Glass Plant Need? | PAVO
A glass plant does not lose power the way a warehouse does. It loses a 1,500°C furnace, a process that runs 24 hours a day, and a restart measured in days, not minutes — if the furnace survives at all.
What size generator does a glass plant need? Most glass plants need 800–4,000 kVA of essential-load backup, not the 2–15 MW the whole plant draws. The set carries the furnace auxiliaries, the forming machines, the annealing lehr and the controls — the loads that keep 500 tonnes of molten glass hot and moving.
In this article:
- Why a Glass Plant Loses More Than Power When the Grid Fails
- What Size Generator Does a Glass Plant Need?
- The Melting Furnace: Flameout, the 1,200°C Line and the Start Sequence
- Forming, Annealing and the Lehr: The Loads After the Furnace
- Fuel, Running Cost and How Long a Glass Plant Can Run
- Heat, Dust and Placement: Siting the Set
- Three Plants, Three Outages
- Frequently Asked Questions
- Specification Checklist
Why a Glass Plant Loses More Than Power When the Grid Fails
Glass is one of the most energy-intensive products on earth. Producing a tonne of glass takes 5–7 GJ of energy, and melting and refining account for 75–85% of the total. Energy makes up roughly 25% of production cost. A float-glass furnace runs 24 hours a day, continuously, for a campaign of 10–15 years, and is stopped only for a rebuild. Heating up and cooling down are long, wasteful and damaging processes. A five-minute stop can cost hours of production, and a bad stop can cost the furnace itself.
That continuity is the point. A float furnace holds up to 500 tonnes of molten glass at 1,350–1,600°C and produces 200–400 tonnes a day — larger fuel-fired furnaces reach 400–700 tonnes a day. The glass inside it never sleeps. It is melted, refined, formed and annealed in one continuous flow, and every stage downstream of the furnace depends on the stage before it staying hot and moving.
A warehouse that loses power loses light and loading. A glass plant that loses power loses the process itself — and, if the furnace cools too far, the furnace.
The furnace is the clock
The melting furnace is a refractory-lined chamber, heated by natural-gas burners with combustion air preheated by the flue gases to as high as 1,300°C in a regenerative furnace. Raw materials — sand, soda ash, limestone and cullet — are fed in at one end. Molten glass leaves at the other, at around 1,500°C, and flows to the forming line. The furnace does not batch. It does not pause. It holds its temperature within a narrow band, day and night, for years.
The danger of a power failure is not that the burners stop. Gas burners can be relit. The danger is what stops with the electricity: the combustion-air fans, the control system, the feeder, the cooling fans on the furnace structure, and the induced-draft fan that holds the furnace pressure. Without them, the furnace pressure goes unstable, cold air is drawn in, and the temperature falls unevenly. Refractory that was designed to sit at a steady 1,500°C is suddenly thermally shocked.
The glass itself is the second problem. Molten glass that stops moving starts cooling, and cooling glass gets more viscous. In a float furnace, the ribbon stops. In a container plant, the forehearths cool and the glass in the feeder channels thickens. In a tempering line, the batch in the furnace is scrapped outright. The restart is not a matter of switching the power back on. It is a matter of bringing a 500-tonne mass of glass back to a uniform, workable temperature without cracking the furnace that contains it.
A furnace rebuild is measured in months and in millions. The French industrial-accident database records a glass-plant furnace refurbishment priced at several million euros, and a full float-furnace rebuild commonly runs into the tens of millions of dollars. No other load on the plant carries that arithmetic. That is why the furnace, not the total plant load, decides the generator.
The forming line stops mid-cycle
The furnace is the most expensive load to lose, but it is not the only one that suffers. When power fails, the forming line stops with glass in it. On a float line, the ribbon of glass on the molten tin bath stops moving, and the tin bath — held at around 1,100°C under a controlled nitrogen-hydrogen atmosphere — starts to cool. On a container line, the individual-section (IS) forming machines stop mid-cycle with hot glass in the moulds. On a tempering line, the glass inside the furnace is heated to near its softening point, about 650°C, and a sudden stop leaves it to warp and deform as the temperature falls unevenly.
The compressed-air system stops too, and that matters as much as the motors. IS forming machines are driven by compressed air at 6–7 bar. Without air, the machines cannot cycle, the moulds cannot open, and the glass in them sets. The annealing lehr — the long, temperature-controlled tunnel that cools formed glass slowly from around 600°C to room temperature — stops its drives and its fans. Glass left standing in a lehr cools too fast, and fast cooling is what creates the internal stress that makes glass crack days or weeks later, in the customer’s hands.
The controls stop too. A glass plant runs on a distributed control system that watches furnace temperature, glass level, forehearth temperature, lehr profile and forming-machine timing. Without power, the operators lose the numbers they need to decide whether the furnace is safe to restart. A plant that restarts blind is a plant that restarts twice — and a glass furnace restarted badly is a furnace that never runs properly again.
What Size Generator Does a Glass Plant Need?
The first mistake in sizing a glass-plant generator is sizing for the whole plant. A 600 tpd float plant draws 8–15 MW of electricity. Backing up all of it would need a 10–18 MVA installation — several containerised sets, a dedicated switchroom and a fuel farm. Almost nobody buys that. The correct question is not how big the plant is, but how much of it must keep running when the grid fails.
The answer is the essential load: the equipment that protects the furnace, keeps the glass moving, and lets the plant restart. The melting itself is mostly gas-fired, which is why the electrical essential load is a modest share of the total. The forming machines, the lehr, the fans and the controls are the loads that matter — and none of them is the batch house, the cullet handling or the packing line, all of which restart after the grid returns.
The essential-load method
An essential-load list for a glass plant has the same shape at every capacity. The items change in size, not in kind:
| Essential load | Typical size | Why it must run |
|---|---|---|
| Furnace auxiliaries: burner control, combustion-air fans, boosting control | 100–300 kW | Keeps the furnace at temperature and under control |
| Forehearths, stirrers and feeders | 30–80 kW | Keeps glass flowing to the forming line |
| Float bath atmosphere and tin-bath controls | 20–50 kW | Protects the tin bath from oxidation |
| IS forming machines and compressed air | 200–600 kW | Keeps moulds cycling and glass moving |
| Annealing lehr drives and fans | 50–150 kW | Cools glass slowly, prevents stress cracks |
| Tempering line drives and quench fans | 100–400 kW | The largest single load on a processing plant |
| Batch-house conveyors and mixers | 50–150 kW | Feeds raw material to the furnace |
| Cooling-water pumps | 50–150 kW | Protects furnace structure, compressors and drives |
| DCS, controls and UPS charging | 20–50 kW | Operators keep their instruments and the restart decision |
| Lighting and communications | 20–40 kW | Safe working during a night outage |
Everything else — cullet crushing, cold-end inspection and packing, the workshop, offices — is shed at the transfer switch and restarted in sequence when the grid returns. The generator sizing guide gives the full method for building a load list and applying diversity and starting factors to it.
Sizing bands by plant type
Applied to real plants, the essential-load method produces a consistent pattern. The figures below are for 50 Hz, 0.8 power factor, with the furnace system fully on the essential board:
| Plant type | Total electrical demand | Essential load | Typical backup set |
|---|---|---|---|
| Small container plant, 100–200 tpd | ~2–4 MW | 400–800 kVA | 800 kVA standby |
| Mid container or float line, 300–500 tpd | ~4–8 MW | 1,000–2,000 kVA | 2,000 kVA standby |
| Large float plant, 600–1,000 tpd | ~8–15 MW | 2,500–4,000 kVA | 2 × 2,000 kVA in parallel |
| Tempering / processing plant, no melter | ~1–3 MW | 300–600 kVA | 500 kVA standby |
Two points in that table deserve attention. First, the essential load is 15–25% of the total plant demand at every capacity — the ratio holds because the furnace auxiliaries and the forming line are a fixed share of the process. Second, above 600 tpd the sensible answer is usually two sets in parallel rather than one large one. Two sets give redundancy on the essential board, let the plant run one set at efficient load during a long outage, and allow maintenance on one set without losing the furnace.
A tempering and processing plant — one that buys flat glass and cuts, tempers and laminates it — is a different and much cheaper problem. There is no melter to protect, no 1,500°C process to save, and the essential load is mostly the tempering line, the cutting tables and the handling equipment. A 500 kVA set covers most processing plants, and the sizing question becomes a straightforward load list.
Motor starting on the forming floor
Sizing by running kW alone will undersize the set. The plant’s motors start hard, and the set must accept the largest single start without the voltage dipping far enough to drop out contactors and drives.
The quench fans on a tempering line are the worst starters on most processing plants: large centrifugal fans, 100–400 kW, started direct-on-line, drawing 5–7 times their running current for a few seconds. The compressed-air compressors are next — 100–300 kW each, and the forming machines cannot run without them. The annealing lehr drives are small by comparison, but they must restart under a full load of hot glass, and a lehr that cannot restart is a lehr full of cracked product.
The practical rule is to restart in sequence, largest motor first, with a 5–10 second gap between starts, and to check the voltage dip against ISO 8528 G2/G3 limits — typically 15–20% transient dip for the loads a glass plant carries. The motor starting article covers locked-rotor kVA, starting methods and voltage dip in full, and the same arithmetic applies here: a set that can run the essential load but cannot start the quench fans is not a backup set, it is a disappointment with a fuel tank.
One more check belongs in the sizing exercise. The furnace boosting electrodes, where fitted, are a large electrical load — sometimes 500–1,000 kW on a float furnace — but they are not an essential load during an outage. Boosting adds heat to the melt; it is switched off when the furnace is in hold mode, and the burners carry the temperature alone. Size the set for the auxiliaries, not the boosters, and confirm the split with the furnace supplier before the order is placed.
The Melting Furnace: Flameout, the 1,200°C Line and the Start Sequence
Everything about the furnace is built around one fact: a hot furnace must stay hot, under control, and under pressure. The burners supply the heat. The fans supply the air. The controls supply the decisions. Lose all three at once, and the furnace is damaged in minutes. Keep the small ones running, and the furnace survives an outage that would otherwise cost millions.
The automatic transfer switch is the first link in that chain. The ATS must detect the failure, start the set and transfer the essential board within 10 seconds — the NFPA 110 window for emergency power — because the furnace emergency procedure starts its clock at the moment of failure, not at the moment the generator reaches speed. A set that starts in 30 seconds is a set that has already spent a third of the safe window doing nothing.

Flameout: the sag that stops the furnace
The most common furnace failure is not a full outage. It is a voltage sag — a dip to 70%, sometimes 50% of nominal voltage, lasting a few cycles, caused by a fault on a parallel feeder or a lightning strike miles away. The sag is too short to start a generator. It is long enough to trip the flame detectors.
A flat-glass plant in the American Midwest learned this the expensive way. Voltage sags from remote faults on the electrical system were causing more than 20 flameouts a year. Each flameout disrupted the process long enough to render the cooled glass unusable, and the scrapped product cost the company hundreds of thousands of dollars a year — close to $1 million annually. Engineers from EPRI’s Power Electronics Applications Center audited the plant and found the cause: the sag made the flame detectors tell the controller, wrongly, that the flame had gone out, and the temperature sensors sent inaccurate readings at the same time. The controller shut the process down on false information.
The fix cost less than $10,000 in power-conditioning hardware. In the year after the audit, the plant had one process stoppage — a transformer failure that caused a half-day outage — and the sag problem was gone. The lesson is not that conditioning replaces a generator. It is that the generator and the conditioning solve different problems: the generator covers the outage, and the conditioning covers the sag that arrives before it. A glass plant that buys a generator without addressing sags will keep losing glass to events the generator never sees.
The 1,200°C line
When the power does fail completely, the furnace procedure is written around temperature. The first action is to stabilise the furnace pressure: adjust the main flue valve to hold the pressure steady, so that excessive negative pressure does not draw cold air into the furnace and drive the temperature down faster. The feeder is raised to avoid burning out, and every part of the kiln is checked while the generator starts.
The critical line is 1,200°C. If the outage lasts long enough for the furnace temperature to fall below 1,200°C, the crown, the tongue arch and the regenerator roof must be inspected for loose wires and refractory movement before restart. Below that temperature, the refractory has contracted enough that the structure is no longer certain, and a restart without inspection risks a roof collapse or a regenerator failure. A furnace that is restarted blind below 1,200°C is a furnace that may not survive the restart.
Glass that cools below its working range in the melter is a separate and worse problem. Molten glass left standing in the furnace thickens, and if it cools far enough it solidifies in place. Removing solidified glass from a furnace means breaking it out by hand, and in the worst case it means a cold repair — a partial rebuild that takes weeks. The French industrial-accident database records a molten-glass spill in which 84 tonnes of glass leaked from a furnace; operating losses were estimated at €700,000 and material damage at €320,000. That was a leak, not a freeze. A frozen furnace costs more.
The generator start sequence
The restart order after a generator start is fixed, and it is worth writing into the emergency procedure rather than leaving to memory. First, the control system, so the operators can see furnace temperature, glass level and pressure again. Second, the primary and secondary air fans, to restore combustion air and bring the melting temperature back. Third, the induced-draft fan, with its frequency adjusted and the flue valve set to hold kiln pressure stable. Fourth, the feeder, to restore the glass level — and if the level has dropped too far, the feeder goes to manual mode, with material added slowly and evenly rather than in a rush. Fifth, the cooling fans at the various parts of the furnace, to protect the structure while it recovers.
Each step has a reason, and the order matters. Starting the feeder before the fans are running pushes cold batch into a furnace that cannot melt it. Starting the induced-draft fan before the control system is up leaves the operators adjusting a pressure they cannot see. The sequence is the same on every glass furnace, and it is the reason the transfer switch must put the control system on the essential board first, before the large motors.
The same logic applies to the cement kiln, which faces an almost identical problem — a hot, continuous process that must keep turning while it cools. The cement kiln article covers the kiln auxiliary drive and the 15-minute warp window in full. The glass furnace has no auxiliary drive to save it. Its protection is the generator, the fans and the procedure — and the procedure only works if the generator is already running.
Forming, Annealing and the Lehr: The Loads After the Furnace
Saving the furnace is the first job. The second job is saving the glass that is already downstream of it. Forming, annealing and tempering are where a power failure turns into scrapped product, and they are where the generator earns its second half.
The float bath and the forming machines
On a float line, molten glass flows from the furnace onto a bath of molten tin at around 1,100°C, where it spreads into a ribbon of controlled thickness and is drawn off by rollers. The tin bath is held under a controlled nitrogen-hydrogen atmosphere. If the power fails, the atmosphere control stops, the bath cools, and the tin oxidises. An oxidised tin bath produces glass with surface defects — the ribbon looks fine at the hot end and is rejected at the cold end, hours later, when the defects are found.
On a container line, the forming machines are the load that matters. Individual-section machines run on compressed air, and each section cycles through blowing, pressing and moulding in a rhythm timed to the second. A power failure stops the machines mid-cycle with hot glass in the moulds. The glass sets in the mould, the mould has to be cleared by hand, and the line restarts with a cold mould that produces underweight, misshapen bottles until it warms up again. A 300 tpd container plant running at full speed loses roughly 12 tonnes of production per hour of stoppage — and the first hour after restart is often lost to warm-up scrap as well.
Compressed air is the hidden dependency. The forming machines need 6–7 bar continuously, and the compressors that supply it are among the largest motors on the essential board. A plant that puts the forming machines on the essential board and forgets the compressors has bought a set that powers machines which cannot move. The compressors, the air receivers and the air dryers belong on the same board, and the receiver must be sized to ride through the 10-second transfer gap without the pressure falling below the forming machines’ minimum.
The annealing lehr
The annealing lehr is the longest and quietest load on the plant, and the easiest to underestimate. It is a tunnel, sometimes 100 metres long, through which formed glass travels on rollers while the temperature falls slowly from around 600°C to room temperature. The slow fall is the whole point: glass cooled too quickly develops internal stress, and stressed glass cracks — sometimes in the warehouse, sometimes in the customer’s hands, weeks after it left the plant.
A lehr that stops with glass inside it cools too fast. The drives stop, the fans stop, and the temperature profile collapses. The glass in the lehr at the moment of failure is almost always scrapped, and the lehr itself must be reheated and re-profiled before production resumes. A lehr restart takes 2–4 hours on a container line, and longer on a float line, where the ribbon must be re-established and the thickness brought back into tolerance.
The lehr drives are small — 50–150 kW — and they are cheap to keep on the essential board. The mistake is to treat them as a minor load and leave them off it. The cost of a lehr full of cracked glass is many times the cost of the power to keep it moving.
The tempering line
A tempering plant has no melter, but it has the most unforgiving load in the glass industry. Tempering heats ordinary glass to near its softening point, about 650°C, and then cools it rapidly and uniformly with high-pressure air, creating compressive stress on the surface and tensile stress inside. The result is glass four to five times stronger than annealed glass — and the process has no tolerance for interruption at any stage.
A sudden power outage during the heating phase stops the process while the glass is hot and unevenly cooled, and the glass warps and deforms from thermal stress. It is scrapped immediately. A stop during the quench phase is worse: the high-pressure fans stop, the glass does not receive the cooling it needs, and the tempering stress never develops. The product is substandard and cannot be sold as tempered glass.
The equipment damage is worse than the product loss. The heating elements — silicon carbide rods or resistance wires — rely on forced cooling by circulating fans at high temperature. After a sudden shutdown, the fans stop but the elements stay extremely hot, and without cooling they overheat, melt or burn out. The ceramic rollers in the furnace operate at high temperature and cool suddenly; the thermal stress bends and deforms them, and a bent roller scratches and distorts every sheet that passes over it afterwards. Replacing a set of ceramic rollers is a slow, expensive job, and the line cannot produce qualified product until it is done. Reputable tempering-furnace manufacturers fit emergency generators and UPS systems as standard, for exactly this reason: the generator’s job is to keep the drives, the circulation fans and the controls running until the glass in the furnace is safely out or safely cooled.
Fuel, Running Cost and How Long a Glass Plant Can Run
A generator that saves the furnace and then runs out of fuel has only postponed the damage. Fuel planning for a glass plant is a different exercise from fuel planning for a hospital or a data centre, because the run is longer and the consumption is larger — and because the furnace cannot be allowed to cool at hour 29 of a 72-hour outage.
The fuel figures below are scaled from the Cummins C500D5 datasheet — 500 kVA standby, 455 kVA prime, 93.2 L/h at full prime load and 69.9 L/h at 75% — and they hold closely enough across the Weichai, Perkins and Cummins ranges that PAVO builds to quote from:
| Set size | Output at 75% load | Fuel at 75% load | Fuel per 24 hours | Fuel per 72 hours |
|---|---|---|---|---|
| 500 kVA | 300 kW | 70 L/h | 1,680 L | 5,040 L |
| 800 kVA | 480 kW | 112 L/h | 2,688 L | 8,064 L |
| 2,000 kVA | 1,200 kW | 280 L/h | 6,720 L | 20,160 L |
| 3,750 kVA | 2,250 kW | 525 L/h | 12,600 L | 37,800 L |
A 2,000 kVA set — the size for a 300–500 tpd plant — burns a 10,000-litre bulk tank in about 36 hours at 75% load. A 3,750 kVA set empties the same tank in 19 hours. The tank that looked generous on the quotation is, in practice, a day and a half of running.

The 72-hour fuel plan
Seventy-two hours is the right planning horizon for a glass plant in a weak-grid market. Grid outages in Nigeria, Pakistan, Bangladesh, Egypt and parts of South-East Asia routinely run to two or three days, and a furnace that has been saved for 48 hours and then lost for want of fuel is a worse outcome than a furnace lost on day one — the plant has burned the fuel, paid the running cost and still faces the refractory bill.
The plan has four parts. First, a bulk tank sized for the longest outage the plant expects, plus a 24-hour margin — 15,000 litres for an 800 kVA set, 25,000 litres for a 2,000 kVA set. Second, a standing fuel delivery contract with a supplier who can reach the plant during an outage, when roads are congested and every other industrial user is ordering at the same time. Third, a low-fuel alarm wired to the control room, with remote monitoring so the plant manager sees it on their phone, set to trigger with at least 12 hours of running left. Fourth, fuel quality management — diesel fuel ageing is a real risk in a tank that sits for months between outages, and a polishing and testing regime belongs in the maintenance schedule alongside the oil and filter changes.
The cost side of the equation is the easiest part of the argument. At $1.00–1.50 a litre, a 2,000 kVA set at 75% load costs $280–420 per hour to run. A 3,750 kVA set costs $525–790 per hour. Set against a furnace rebuild measured in millions of dollars, and a 300 tpd container plant losing roughly 12 tonnes of production per hour of stoppage, the fuel bill is a rounding error. The expensive decision is not running the generator. The expensive decision is not having one, or having one with an empty tank.
One rating question belongs here. If the plant’s grid is weak enough that the set runs 1,000 hours or more a year, it is a prime-power set, not a standby set. The standby and prime power ratings decide warranty, service intervals and design life, and a standby-rated set run as a prime set will follow the familiar pattern: acceptable for months, then rising oil consumption, then an early overhaul and a refused warranty claim. Glass plants in weak-grid markets are prime-duty sites by definition, and the quotation should say so.
Heat, Dust and Placement: Siting the Set
A glass plant is one of the hardest environments a diesel generator will ever work in. The air is hot and full of fine, abrasive dust. The ambient temperature in the Middle East, North Africa and South Asia regularly exceeds 45°C. The set is expected to start, cold, after weeks idle, and take full load in seconds. Placement and protection decide whether it can.
Dust and the air filter
Glass-plant dust is not ordinary dirt. Batch-house dust is fine silica and soda ash — abrasive enough to wear cylinder bores if it reaches the combustion air, and alkaline enough to attack metal surfaces. A set installed in a glass plant with a standard single-stage filter will show rising fuel consumption and falling power within months, and the failure will be blamed on the engine rather than on the air it breathes.
The fix is layered. A cyclone pre-cleaner or a two-stage air filter keeps the bulk of the dust out of the element. The element itself is inspected weekly and changed on condition, not on the calendar — a filter that looks clean after a month in a packing hall is a filter that is not doing its job, and a filter that is black after a week near the batch house is a filter that needs changing now. The radiator is washed down on the same weekly round, because a radiator blocked with dust is a set that overheats at 60% load on a hot day.
The enclosure matters too. An open set in a glass plant needs a roof, a bund and a clean-air intake. A silent canopy set is the better default: the canopy keeps dust off the engine and the alternator, and its intake can be ducted away from the dustiest part of the plant. The canopy is rated IP54 or better, and the intake is pointed away from the batch house, the cullet crusher and the furnace charging end — the three dustiest points on the site.
Heat, derating and ventilation
The second environmental fact is heat. Ratings are stated at 25°C and sea level. A glass plant in Saudi Arabia, Egypt or Pakistan sees 45–50°C for months of the year, and the set loses output whether the budget allows for it or not. The alternator loses about 3% for every 10°C above 40°C. A set rated 2,000 kVA at 25°C delivers roughly 1,850 kVA at 50°C — and that is before the radiator, sized for the same conditions, is asked to reject full-load heat into 50°C air.
The remedy is to specify the set for the site, not the catalogue. Ask for the derated figure at the actual maximum ambient temperature, and size the radiator for 50°C ambient with a margin. Generator room ventilation is the other half of the same calculation: the room needs enough air for combustion and for radiator cooling, and in a glass plant that air must come through a filtered louvre, not an open vent that admits dust by the tonne.
Placement ties the two together. The set should sit close to the furnace electrical room, to keep cable runs short and voltage drop low, but away from the batch house and the cullet crusher. It should sit on a concrete plinth above the flood line, with a bund holding 110% of the fuel tank volume, and with clear access on all four sides for the weekly filter and radiator round. A set that cannot be reached for maintenance will not be maintained, and a set that is not maintained will not start when the furnace needs it.
Three Plants, Three Outages
The rules above are not theory. Three plants, in three countries, learned them in the order that matters: one by measurement, one by preparation, and one by paying for the lesson.
Case 1: A flat-glass plant in the American Midwest — the sags that cost a million dollars a year
A flat-glass plant in the American Midwest was losing glass to power events that never reached the level of an outage. Voltage sags from remote faults on the electrical system were causing more than 20 flameouts a year. Each flameout disrupted the furnace process long enough to render the cooled glass unusable, and the scrapped product cost the company hundreds of thousands of dollars a year — close to $1 million annually.
Engineers from EPRI’s Power Electronics Applications Center audited the plant and found the cause in the control chain, not in the furnace. The sag made the flame detectors tell the controller, wrongly, that the flame had gone out. The temperature sensors sent inaccurate readings at the same time. The controller shut the process down on false information, and the furnace lost a batch of glass to an event that lasted a fraction of a second.
The fix cost less than $10,000 in power-conditioning hardware. In the year after the audit, the plant had one process stoppage — a transformer failure that caused a half-day outage — and the sag problem was gone. The plant still needed its generator for the real outage. What it learned was that the generator and the conditioning are two halves of the same system: the generator covers the outage, and the conditioning covers the sag that arrives before it. A glass plant that buys one without the other keeps losing glass to the half it forgot.
Case 2: A 300 tpd container plant in Egypt — the melter that stayed hot
A 300 tpd container-glass plant in Egypt runs on a grid that fails two or three times a month, usually for 4–12 hours. The plant’s essential board is carried by a 2,000 kVA Cummins set, prime-rated, with a 25,000-litre bulk tank and a 10-second automatic transfer switch.
At 01:40 on a Thursday in the hot season, the grid failed with the furnace at full production. The transfer switch started the set and transferred the essential board in 9 seconds. The combustion-air fans held, the feeder kept the glass level steady, and the operators held the furnace pressure on the flue valve while the set settled. The forming machines kept cycling on compressed air, and the lehr kept its temperature profile.
The grid returned 8 hours later. The furnace had never fallen below 1,350°C. The forming machines restarted from warm moulds, the lehr needed no re-profiling, and the plant was back at full production the same shift. No refractory damage. No frozen glass. No cold repair.
The same plant had lost a furnace crown in 2020, before the backup set was installed. A 35-minute outage let the temperature fall below 1,200°C, the crown wires loosened, and the restart cracked the refractory. The repair cost $240,000 with 9 days of lost production. The 2,000 kVA set, the transfer switch and the bulk tank cost less than that single failure.
Case 3: A tempering plant in the UAE — the quench fans that stopped mid-cycle
A tempering plant in the United Arab Emirates processed architectural glass for high-rise projects, running two tempering lines on a grid that was reliable enough that backup power had never been a priority. The plant had a small generator for the offices and the lighting. It had nothing for the lines.
The outage came at 14:20 on a summer afternoon, with both lines loaded. The heating phase stopped with glass at 650°C inside the furnace. The quench fans stopped with it. The glass in the first line warped as it cooled unevenly and was scrapped. The glass in the second line never developed its tempering stress and was rejected as substandard.
The equipment damage was worse. The silicon carbide heating elements, left at high temperature without their cooling fans, overheated and burned out — a $35,000 replacement. The ceramic rollers deformed from the sudden thermal shock, and every sheet that passed over them afterwards carried scratches and optical distortion. The line was down for 6 days while the rollers were replaced and the furnace re-qualified, and the plant missed the delivery dates on two tower contracts.
The fix was a 500 kVA set on the essential board, carrying the drives, the circulation fans and the controls, with a UPS bridging the 10-second transfer gap. The set cost less than the roller replacement alone. The lesson is that a tempering line without backup power is not a tempering line with a risk. It is a tempering line with a countdown.
Frequently Asked Questions
What size generator does a glass plant need?
Most glass plants need 800–4,000 kVA of essential-load backup, depending on plant type. A small container plant of 100–200 tpd typically needs an 800 kVA set, a 300–500 tpd float or container line needs 2,000 kVA, and a large float plant of 600–1,000 tpd needs two 2,000 kVA sets in parallel. The correct figure always comes from an essential-load list — furnace auxiliaries, forming machines, lehr and controls — never from the plant’s total 2–15 MW demand.
Do glass factories need a backup generator?
Yes, in almost every case. A glass plant loses money within minutes of a power cut: a furnace that falls below 1,200°C needs its crown and regenerator inspected before restart, glass left standing in the melter thickens and can solidify, and a tempering line loses its batch and its rollers in a single stop. In countries with unreliable grids, such as Nigeria, Pakistan, Egypt and Bangladesh, a plant without a generator risks its furnace on every outage, and buyers will move orders to plants that can guarantee delivery.
How long can a glass plant run on a generator?
As long as the fuel lasts. A prime-rated set can run indefinitely, limited only by fuel supply and the 250-hour service interval. A 2,000 kVA set at 75% load burns roughly 280 litres an hour, so a 10,000-litre bulk tank gives about 36 hours of running. A plant should size its tank for the longest outage it expects, plus a 24-hour margin, and keep a standing fuel delivery contract in place.
What happens to a glass furnace during a power outage?
The burners stop, but the dangerous part is what stops with the electricity: the combustion-air fans, the control system, the feeder and the induced-draft fan. The furnace pressure goes unstable, cold air is drawn in, and the temperature falls unevenly. If the temperature falls below 1,200°C, the crown and regenerator roof must be inspected before restart, and glass left standing in the melter thickens and can solidify. The emergency procedure is to stabilise the furnace pressure on the flue valve, then restart in order — controls, air fans, induced-draft fan, feeder, cooling fans — once the generator is running.
Can a diesel generator power a glass factory?
Yes — the essential loads, which is what matters. Diesel generators from 500 kVA to 4,000 kVA power glass plants across Africa, South Asia and the Middle East every day, and they are the standard backup source for the industry. Backing up the whole plant would need 10–18 MVA, which almost nobody buys; the set carries the furnace auxiliaries, the forming machines, the lehr and the controls, while the batch house and the packing lines restart when the grid returns.
Specification checklist — before the order is placed
- Load list complete: furnace auxiliaries, forehearths, forming machines, compressors, lehr, tempering line, cooling water, controls — listed with running kW, starting current and duty
- Rating checked: prime-rated if the set will run more than 200 hours a year, standby-rated only for genuinely occasional outages
- Furnace protection proven: emergency procedure written, 1,200°C inspection rule in it, restart sequence — controls, air fans, induced-draft fan, feeder, cooling fans — rehearsed under load
- Starting surge verified: staggered restart programmed, soft starters or variable-speed drives on large fans and compressors, largest motor start within voltage dip limits
- Fuel autonomy confirmed: bulk tank sized for the longest expected outage plus 24 hours, low-fuel alarm fitted, delivery contract in place, fuel tested
- Dust plan in place: cyclone pre-cleaner, weekly filter and radiator cleaning, intake away from the batch house, IP54 enclosure or silent canopy
- Testing regime agreed: monthly load test with the full restart sequence, not a no-load run, with the furnace procedure and the transfer switch proven every time
A glass plant buys a generator once and lives with it for 15–20 years. The plants that get it right size from an essential-load list, buy a prime-rated set, put the furnace system on the essential board, and test the whole system under load every month. The plants that get it wrong buy from the total plant load, discover the furnace on the first real outage, and pay for the lesson in refractory and lost production. To size a set for a container line, a float plant or a tempering line, contact PAVO with the load list, and the engineering team will return a sized quotation with the fuel, tank and transfer switch specified to match.