Transformer fire protection systems: designing deluge and water spray to NFPA 15, with real figures and the Thai legal position

How much risk does an oil-filled transformer carry, and when does it need a fixed fire system

A mineral-oil-filled power transformer is a steel tank holding tens of thousands of litres of combustible liquid, standing in the open, under hundreds of thousands of volts. Thai training material gives the fire point of flammable transformer oil as 165 °C, while an insulating liquid classed as “less flammable” must have a fire point of 300 °C or above. That pair of numbers is the whole reason oil-filled transformers are treated differently from dry types.

The frequency of events is low but not zero. The IEEE technical paper PCIC 2022-PCIC-0545 cites CIGRE WG A2.33 for a major failure rate of roughly 1 per cent per transformer per service year, with roughly one in ten severe failures ending in fire, for transformers of 10 MVA and 66 kV and above. The same paper gives a cumulative figure of about 4 per cent over a 40-year service life, and reports IEEE Std 979 Table A.2 as giving fire frequencies of 0.025 to 0.09 per cent per year across the 69 kV to 500 kV range. All of these figures are reported here through that IEEE PCIC paper — we have not read the CIGRE report or IEEE 979 directly — so they should be read as orders of magnitude, not as values to calculate with.

A fire system is the third option, not the first

The sharpest statement on this sits in NFPA 70 (NEC) 450.27, quoted in full in the City of Phoenix guidance document. Where an oil-filled transformer installation presents a fire hazard, one or more safeguards shall be applied according to the degree of hazard, in this order: (1) space separation, (2) fire-resistant barriers, (3) automatic fire suppression systems, and (4) enclosures that confine the oil of a ruptured transformer tank.

That means a deluge system is not the default on every project. It is what gets installed when the site layout will not allow separation or a barrier, or when the value and criticality of that particular transformer make total loss unacceptable. The first design question is therefore not “how many nozzles” but “can the transformer yard layout still be changed?”

Thai law on transformer fire systems: what is mandatory, and what is barely addressed at all

This is where most published material in this market becomes vague, so it is worth being exact.

For an outdoor transformer yard, no Thai instrument mandates a fixed fire-extinguishing system. There is no equivalent of the LPG cooling-water notification or the oil depot foam requirements. What Thai law does provide is a floor for firewater duration and a set of conditions that apply once you have chosen to install a system.

Where Thai law does bite

• กฎกระทรวง ฉบับที่ 33 (พ.ศ. 2535) ข้อ 20 applies to transformer rooms inside high-rise and extra-large buildings — an indoor requirement, not an outdoor yard one.
• ประกาศกระทรวงอุตสาหกรรม พ.ศ. 2552 ข้อ 10 sets a firewater supply of not less than 30 minutes’ continuous duration.
• กฎกระทรวงแรงงาน พ.ศ. 2555, Table 1 sets a stored firewater reserve of 9,000 to 36,000 litres according to building area.
• มอก. 2541 เล่ม 8 is the Thai standard covering water spray fixed systems, and it is the document that gives Thai names to the concepts: a water spray system is “ระบบฉีดน้ำฝอยดับเพลิง” and a deluge system is formally “ระบบเปิดทะลัก” under clause 7.2.1.2.

The conclusion you can act on: for an outdoor transformer in Thailand the design basis comes from international standards — NFPA 15 and NFPA 850 — together with the requirements of the insurer and the asset owner. Thai law contributes a firewater duration floor and some installation conditions, and nothing more. Anyone telling you that Thai law requires a deluge system on an outdoor transformer should be asked which clause.

A note on NFPA 850’s status, which has changed. From the 2026 edition, effective 9 December 2025, NFPA 850 was elevated from a Recommended Practice to a Standard. The title changed from “Recommended Practice for Fire Protection for Electric Generating Plants…” to “Standard for Fire Protection for Electric Generating Plants and High Voltage Direct Current (HVDC) Converter Stations”. Requirements moved into the body of the document, informational guidance moved to the annex, and a Fire Protection Design Basis Document (FPDBD) became mandatory. Two consequences follow: clause numbers may have moved from the 2020 edition cited later on this page, and wording that was “should” may now be “shall”, which changes the contractual status of the provision. We have not read the 2026 text ourselves — this comes from NFPA’s own announcement and from standards-vendor listings, not from the text — so we quote no 2026 clause numbers. Check against the edition in force before using it in a design or a TOR.

Outdoor transformer yards and indoor transformer rooms: different hazards, different systems, do not mix them

Search for transformer fire protection in Thai and the results always conflate two things: systems for an outdoor transformer yard and systems for an indoor transformer or electrical room. They are not the same problem.

Inert gas and clean agent systems are not the answer for an outdoor transformer, because there is no building envelope to hold the agent concentration and they give no cooling effect on hot steel. Several machine-translated Thai pages propose Argonite or Inergen for “substations”, which takes a control-room system and offers it as the answer to an outdoor yard problem.

Indoor figures worth knowing for comparison

Thai training material reproducing clause 6.4(c) of the EIT standard gives transformer room walls and roof as reinforced concrete not less than 125 mm thick, or brick or concrete block not less than 200 mm — with a note that where the installation has automatic fire-extinguishing equipment, those thicknesses reduce to 65 mm and 100 mm respectively. This is one of the few places where a Thai document puts a number on what a fire system is worth.

The same material gives clearances of not less than 1.00 m between a transformer and the room wall or door, not less than 0.60 m between transformers, not less than 0.60 m above, and a door sill not less than 0.10 m high to contain oil in the event of a rupture. And the sentence that matters most for deluge work: no unrelated piping may be located in or pass through a transformer room, except piping for fire protection or for transformer cooling. That sentence disposes of the most common objection raised in Thai tenders, that firewater piping cannot be run into a transformer room.

For outdoor yards the same material cites clauses 6.4.1 and 6.4.2: live parts of a high-voltage system above a working space must be not less than 2.75 m above floor level or be guarded; horizontal clearance between a fence or wall and live high-voltage parts not less than 1.20 m for voltages up to 33 kV; and between fence and transformer not less than 1.00 m.

Two cautions. First, these figures were read from training material reproducing the standard rather than from the printed EIT 022001-22 volume, and the version reproduced appears to be the 2556 (2013) edition — so verify against the current volume before writing them into a specification. Second, a great many Thai articles misprint the 2.75 m height as “2.75 centimetres”. Do not copy that on.

What NFPA 15 is, and which clause applies to transformers

NFPA 15, Standard for Water Spray Fixed Systems for Fire Protection, is the American standard governing the design of fixed water spray systems, covering extinguishment, control of burning, exposure protection and prevention of fire. It is the document that says how much water, at what pressure, distributed how, and how fast the system must respond. Thailand’s มอก. 2541 เล่ม 8 calls this type of system “ระบบฉีดน้ำฝอยดับเพลิง” and the nozzle a “หัวฉีดกระจายน้ำฝอย (water spray nozzle)”.

A deluge system is a type of piping arrangement, not a type of nozzle. มอก. 2541 เล่ม 8 clause 7.2.1.2 gives it the official Thai name “ระบบเปิดทะลัก” and defines it as a system in which open nozzles connect to piping fed through a control valve, with no water in the pipe or nozzles in the normal state; on detection, the valve opens and water discharges simultaneously from every nozzle. In practice nobody in the Thai market writes “ระบบเปิดทะลัก” — everyone writes ระบบเดลูจ or Deluge System, and calls the valve a Deluge Valve.

The transformer clause, and the clause-numbering history that causes misquotation

We compared two editions of NFPA 15 twenty-seven years apart, line by line, and found the transformer requirements essentially unchanged. What changed is the clause number — which is why engineers so often cite the wrong one.

The text we read word for word comes from the 1996 and 1969 editions, the latter published in a US government agency reading room. The clause number 7.4.4 in the current edition is secondary information — we did not open the current text ourselves. That said, NFPA’s own technical committee agenda documents reprint current clause 7.4.2 (tanks and vessels) in full, and it matches 1996 clause 4-5.2 word for word including the 0.25 gpm/ft² density, which is strong evidence that chapter 7 carried chapter 4’s text across intact.

Recommendation for a TOR: cite it as “NFPA 15 clause 7.4.4 (edition in force at the date of tender)” and do not copy clause numbers like “4-5.4.2” out of older textbooks, because the contractor will not find them in the volume he has bought.

The most common misquotation: “NFPA 15 requires a 15-minute minimum”

This is wrong for transformer work. The 15-minute figure in the standard sits at clause 9-3.5 of the 1996 edition, inside chapter 9, Ultra High-Speed Water Spray Systems — systems for explosion hazards requiring discharge within fractions of a second. It is not the transformer deluge provision. Anyone writing 15 minutes into a transformer specification is citing the wrong chapter.

Water density for transformers: 10.2 and 6.1 L/min/m², and the area basis nearly everyone gets wrong

NFPA 15 specifies two densities for a transformer, not one:

• 10.2 L/min/m² (0.25 gpm/ft²) over the projected area of the rectangular prism envelope enclosing the transformer and its accessories.
• 6.1 L/min/m² (0.15 gpm/ft²) over the nonabsorbing ground surface area expected to be affected.

Almost all Thai-language material currently available cites only the first and drops the second, which systematically undersizes the system — because the fire that actually threatens the transformer tank is the oil fire pooling underneath it, not the flame at the tank surface.

The “rectangular prism envelope” is where the calculation most often breaks

The standard does not have you multiply the density by the transformer’s true surface area. It has you multiply by the projected area of the rectangular prism envelope enclosing the transformer and all its accessories.

The difference is enormous. Calculate the true surface area of every radiator fin and the area inflates several times over, the flow rate becomes wildly excessive and the pump is oversized. Calculate only the box of the main tank, excluding the radiator bank and the conservator, and the area is too small and the system runs short of water. The correct method is to draw a rectangular box around everything that projects, and use the projected area of that box.

Does this density change between editions?

No — at least not for the transformer clause. We compared the 1969 and 1996 editions and got the same pair of values, and an article by an NFPA technical services engineer published on nfpa.org on 29 October 2021 gives the same pair for the edition then in force. (That article carries NFPA’s standard disclaimer that it represents the author’s views, so it should be cited as an article, not as the standard.)

An independent source giving the same figure

The property risk guidance of AXA XL (PRC.12.21.2) calls for oil-filled transformers critical to production to be protected by water spray impinging on all exposed exterior surfaces at a density of 0.25 gpm/ft² (10.2 L/min/m²), and its conversion table gives 1 gpm/ft² = 40.7 L/min/m². That is a third source, outside NFPA, giving the same value.

A higher figure that genuinely exists

UFC 3-600-01, the US Department of Defense facilities design criteria, edition of 8 August 2016 Change 6, sets a higher figure: density 0.30 gpm/ft² (12.2 L/min/m²) over the transformer surfaces excluding the area beneath the transformer, with a water supply sufficient for 2 hours and including a hose stream demand of 500 gpm (1,900 L/min). This document is freely available and we read the text directly.

Do not present any single figure as “the correct” one. 0.25 gpm/ft² is a minimum, not a universal answer. An owner or insurer may require more, and if the TOR does not state which basis applies, each bidder will calculate on a different one and the prices cannot be compared.

The 10.2 figure already contains a 2.0 wastage allowance — do not add a second safety factor

This is information almost never published in Thai, and it saves real money.

Annex A-4-5.1(b) of the 1996 edition states that the densities specified for exposure protection already include a 0.05 gpm/ft², or 2.0 L/min/m², allowance to compensate for unexpected water losses, with the standard defining “wastage” as water lost to factors such as wind effect and nozzle discharge angle tolerance. The 1969 edition says the same thing — that the standard anticipates a minimum wastage of 0.05 gpm per square foot — and adds that in some cases a greater allowance should be made.

In working numbers: of the 10.2 L/min/m², roughly 8.2 is water doing work and 2.0 is the allowance.

What follows for design

Many engineers multiply 10.2 by their own safety factor of 1.2 or 1.3 “to allow for wind”. That is double counting, because wind is precisely what the 2.0 was added to compensate for. The result is a larger pump, a larger tank and larger mains, without a proportionate increase in fire-fighting capability.

What to do instead is to be explicit about what you are allowing for. If the site is unusually windy, address it through nozzle discharge angle and mounting distance and through working pressure — Annex A-5-1.2 says itself that the minimum pressure exists to let the discharge pattern develop fully and to overcome wind effects — rather than by raising the density across the whole system.

And if you do decide to add margin, write into the calculation how much and why, so that a reviewer can see it as an engineering decision rather than a number that arrived from nowhere.

‘Nonabsorbing ground’ is defined more broadly than people assume, and it widens the 6.1 area considerably

A common misunderstanding is that the 6.1 L/min/m² density applies only to the concrete floor of the oil containment pit, and that the surrounding ground need not be counted.

The chapter 1 definition in the 1996 edition settles it. Nonabsorbing ground means earth or fill that will not readily permit large quantities of flammable liquid or water to penetrate or be absorbed, and the standard goes on to say that most soils are not considered sufficiently permeable or absorbent to qualify as absorbing ground. Hard surfaces such as concrete or asphalt are nonabsorbing in any case.

By default, therefore, nearly the whole area of a transformer yard qualifies as nonabsorbing ground, and the 6.1 figure must cover the ground surface area expected to be affected, not merely the rim of the pit.

How to define that boundary defensibly

The standard’s phrase is “expected to be affected”, which is a designer’s judgement rather than a fixed number. Defensible practice is to:

• set the boundary from the shape of the containment area and the actual slope of the ground, not from the edge of the transformer’s concrete plinth;
• treat ground that oil would cross before reaching a drain or channel as affected area;
• record that assumption in writing in the calculation, because it is the first thing a reviewer or an insurer will ask about.

Assuming too small an area is the easiest way to reduce the flow rate on paper — and the easiest way to make the system fail in reality.

The most counter-intuitive clause in NFPA 15: do not spray energized bushings and surge arresters directly

Everyone assumes the principle is to cover everything. For energized components the standard says the opposite.

NFPA 15 clause 7.4.4.4 (formerly 4-5.4.4) requires nozzles to be positioned so that the water spray does not envelop energized bushings or surge arresters by direct impingement. There is one exception: where the transformer manufacturer or its documentation permits it, and the owner permits it. The 1969 edition, clause 4043(d)(3), states the reason plainly — to prevent damage to energized bushings and surge arresters.

We read both texts word for word. During research we found a vendor page describing “dedicated nozzles for HV and LV bushings” as standard practice, citing no clause at all. That is doing exactly what the standard prohibits, absent written permission from the manufacturer and the owner.

Why the standard is afraid of water at the bushing

The engineering explanation is consistent with what FIST 3-32 of the US Bureau of Reclamation records: firewater contains impurities which, applied continuously, damage a transformer’s external components and can cause surface flashover. The porcelain surface of a bushing, wetted with conductive water, is a flashover path.

What about surfaces the spray cannot reach?

The standard answers that too. Clause 7.4.4.1 (formerly 4-5.4.1) requires water spray to impinge on all exposed exterior surfaces, but Exception 1 permits that where there is insufficient space to place nozzles beneath the transformer to impinge directly on the underside, those surfaces may be protected by horizontal projection or by nozzles directed to cool the area beneath the transformer’s projecting parts.

Note what the standard does not permit. It never allows credit for water running down surfaces for transformers, and for tanks and vessels clause 4-5.2.4 explicitly forbids assuming that surfaces below the equator of a tank are wetted by rundown. The standard is consistently sceptical of rundown credit.

Nozzle layout and piping: the 305 mm gap, the conservator, 3 m spacing and electrical clearance

Once the density is fixed, what decides whether the system actually works is the nozzle layout. NFPA 15 contains several geometric rules that can be checked directly off a drawing.

Gaps wider than 305 mm need separate protection

Clause 7.4.4.2.2 (formerly 4-5.4.2.2) states that where transformer components create gaps exceeding 12 in (305 mm) in width, those surfaces must be individually protected. The 1969 edition, clause 4043(d)(1), gives the same twelve-inch criterion for gaps between radiator panels.

This is a genuinely checkable rule for radiator banks and cooler assemblies. If the spacing between fin panels exceeds 305 mm, you cannot treat the assembly as a single solid face.

The conservator is not an optional extra

Clause 7.4.4.2.1 (formerly 4-5.4.2.1) applies the same density to special arrangements, the conservator tank and pumps. The 1969 edition lists the same items. The conservator sits high and out of the natural line of sight, so it is routinely forgotten in nozzle layouts — despite being full of oil.

No piping across the tank top or the control cabinet face

Clause 7.4.4.3 (formerly 4-5.4.3) prohibits running water spray piping across the top of the transformer tank or across the face of its control cabinet. The exception applies only where impingement cannot be achieved by any other arrangement, and full electrical clearances from live parts must still be maintained. This constrains the piping layout directly and is almost never mentioned in marketing content.

Maximum nozzle spacing

Clauses 7.1.7 and 7.1.8 (formerly 4-1.4) require nozzle spacing, both vertical and horizontal, not to exceed 10 ft (3 m) unless the nozzle is listed for greater spacing, and discharge patterns must at least meet. A system that satisfies the density on paper can still fail this clause if it uses too few, too large nozzles.

Clearance from uninsulated energized parts

Table 3-1.2 of the 1996 edition gives minimum clearances from water spray system equipment to uninsulated energized parts. The values matching voltage levels used in Thailand are:

This is the physical reason you cannot wrap pipework around a transformer as you please. At 230 kV the pipe and nozzles must stand back nearly two metres from live parts, which immediately affects the discharge angle and throw required. Electrical clearance and nozzle layout have to be designed together, not one after the other.

Nozzle pressure, K-factor, and the unit trap that puts a system out by a factor of 14

Minimum pressure at the nozzle

Clause 8.1.2 (formerly 5-1.2) sets a minimum of 20 psi (1.4 bar) at any nozzle protecting an outdoor hazard, which is what a transformer in a yard is. Nozzles protecting indoor hazards follow their own listing.

Annex A-5-1.2 explains that the minimum pressure exists so that the discharge pattern develops properly and to overcome wind effects, and recommends 30 psi for nozzles with orifices of 3/8 in or smaller. For an outdoor yard in Thailand, with both monsoon winds and summer storms, that recommendation carries more weight than it does in the original.

A warning about a typographical error in the source. The 1996 annex prints the metric equivalent of 30 psi as “(1.4 kPa)”, which is wrong by three orders of magnitude. The correct value is approximately 2.07 bar, or 207 kPa. Do not copy the printed conversion.

The nozzle flow formula and the 14.4 factor

Clause 8.5.1.5 (formerly 5-5.1.5) gives the nozzle flow formula in two unit systems, and this is where systems go badly wrong. The imperial form uses Q in gpm and P in psi; the metric form uses Q in L/min and P in bar, and the K value is not the same number in the two systems. The metric K is the imperial K multiplied by approximately 14.4.

Take a K value from a US catalogue, put it into a metric formula with pressure in bar, and the calculated flow is out by that factor. It is the single most destructive arithmetic error in this discipline, because it produces a number that looks plausible on a spreadsheet. Always state which unit system the K value belongs to, in the calculation and in the TOR.

Water quantity and duration: how far apart the Thai and international figures are

The NFPA 15 minimum

Clause 7.4.4.2.3 (formerly 4-5.4.2.3) requires the water supply to deliver the design flow rate plus 250 gpm (946 L/min) for hose streams for not less than 1 hour. We verified this from the 1996 text directly, and an article by an NFPA technical services engineer on nfpa.org in 2021 gives the same set for the edition then in force.

But one hour is a minimum, not a target. Clause 7.4.1 (formerly 4-5.1.1) sets the principle that an exposure protection system must perform as intended for the duration of the expected exposure fire, and Annex A-4-5.1.1 directs the designer to weigh the properties and quantity of the fuel providing the exposure, and the effect of manual fire-fighting, noting that operation for several hours may be required.

This is the layer where judgement enters. A system delivering for one hour may be standard-compliant and still inadequate in practice for a transformer holding tens of thousands of litres of oil.

The three layers of Thai figures, set against the international ones

มอก. 2541 เล่ม 8 clause 4.2.3.2 defines “extra hazard group 2” as an area with large quantities of combustible material, or with flammable liquids or gas present in quantity — and a transformer yard holding tens of thousands of litres of oil has a strong case for that classification, which carries 120 minutes. That said, hazard classification is a designer’s judgement, not an automatic result of the text, and the class chosen must be stated in the TOR. It aligns far better with UFC than with the 30-minute Thai statutory floor.

A caution on document status: the มอก. 2541 เล่ม 8-2560 file published on the TISI website carries a watermark on every page reading “ใช้ในการรับฟังความคิดเห็นเท่านั้น” — for consultation only — even though the header carries the standard number and year. We found no Royal Gazette notice confirming promulgation of a final version. Cite it with its status stated, or check with TISI before making it binding in a TOR.

Provisions from มอก. you can write into a specification immediately

มอก. 2541 เล่ม 8 clause 7.4.1(1) requires water spray nozzles to be tested and listed by a recognised body such as UL or FM Global. Clause 7.4.2(2) requires the control valve to be installed outside the protected area and safe from fire. Clause 7.4.3 requires detection to NFPA 72 or EIT 2002, and clause 7.4.4 requires detector heads to EIT 3002 and NFPA 13, with spare nozzles specified as: fewer than 300 nozzles, at least 6 spares; 300 to 1,000, 12 spares; more than 1,000, 24 spares.

The UL/FM requirement at national standard level is a useful procurement tool, because it excludes unlisted nozzles from consideration without having to rely solely on a foreign standard.

Indicative worked example: from transformer size to water tank and oil pit

Important warning: every figure in this section is indicative, to show the method. All inputs are values we assumed ourselves. They come from no project, and they do not substitute for a real calculation certified by a licensed engineer.

Assumed inputs

Step 1 — prism envelope area

• Two long faces: 2 × (4.0 × 4.5) = 36.0 m²
• Two end faces: 2 × (3.2 × 4.5) = 28.8 m²
• Top: 4.0 × 3.2 = 12.8 m²

Total 77.6 m²

Step 2 — flow on the transformer

77.6 m² × 10.2 L/min/m² = 791 L/min

Step 3 — flow on the ground

31.2 m² × 6.1 L/min/m² = 190 L/min

Step 4 — total design flow

791 + 190 = 981 L/min. Note that the ground portion is roughly one fifth of the total. That is the quantity missing from every Thai page that cites only the 10.2 figure.

Step 5 — approximate nozzle count

• Flow per nozzle at 2.0 bar: Qm = 28.8 × √2.0 = 40.7 L/min per nozzle
• 981 ÷ 40.7 = approximately 25 nozzles

This is only a starting point. The real layout must be checked again against three constraints: spacing not exceeding 3 m, gaps over 305 mm requiring individual protection, and electrical clearance by voltage. In practice the real nozzle count is usually higher than the flow calculation alone suggests, and once the count rises the total flow rises above 981 L/min — and it is that higher figure, not the minimum density figure, that must be used.

Step 6 — total water demand

Design rate 981 + NFPA 15 hose stream 946 = 1,927 L/min

• At 60 minutes (the NFPA 15 minimum): 1,927 × 60 = 115,620 litres, about 116 m³
• At 120 minutes (extra hazard group 2 under มอก. 2541 เล่ม 8): 231 m³

Set that against the maximum stored reserve in Table 1 of กฎกระทรวงแรงงาน 2555, which is 36,000 litres: the entire Thai statutory reserve is consumed in under 20 minutes by this one system. A transformer fire system therefore needs its own tank and its own pump. It cannot share the building’s water.

Step 7 — oil pit volume

FIST 3-32 sets the principle that containment volume must hold the oil that may be released, plus the firewater expected to be discharged, plus accumulated rainwater.

• Oil: 12,000 litres
• Water from the spray system over 60 minutes, assuming nearly all of it reaches the containment area: 981 × 60 = 58,860 litres
• Water from hose streams and supplementary monitors over 60 minutes: 946 × 60 = 56,760 litres. Clause 4-2.4 item (2) identifies hose streams likely to be used as a component to be counted, and Step 6 has already calculated this figure, so it must be carried into the pit as well.
• Rainwater, assuming 100 mm/hr over 31.2 m²: 3,120 litres

Total approximately 130,700 litres, or about 131 m³

Hose stream water is the variable a designer must state an assumption about, because the direction of application determines how much reaches the pit. Counting only the spray system and excluding hose streams entirely gives about 74 m³. That range, 74 to 131 m³, is exactly what a TOR has to close — otherwise each bidder designs a different pit and none of them is wrong.

What this set of figures shows clearly is that firewater, not oil, is the largest part of the containment volume. The 12,000 litres of oil is less than a tenth of the total once hose streams are included. That is why installing a deluge system over an existing containment pit without enlarging it converts a fire incident into an oil spill incident.

If the 120-minute duration is selected, the pit figure doubles again, which is often physically impossible. In that case the answer is controlled discharge, not an ever-larger pit — see the detection section below.

The transformer oil pit: volume, stone fill, and the rain problem Thailand has to add

A question that has sat on Thai engineering forums for years is how to size a transformer oil pit. The answer in circulation is three times the oil volume — which is half right.

The Thai “three times” rule, and its actual scope

Training material reproducing clause 6.4(c) of the EIT standard states that an oil sump must hold not less than three times the liquid volume of the largest transformer, that it should be filled with No. 2 stone, and gives the formula C_SUMP ≥ 3 × V_FLUID MAX. Where the sump is outside the room, a pipe of not less than 50 mm must drain liquid from the transformer room to the sump, with the transformer-side end screened. The worked example given is a 1250 kVA transformer holding 860 litres of oil, requiring a 2,580 litre sump.

Three times is a demanding requirement, but it must be read in its actual scope: it is a rule for transformer rooms, not for outdoor yards. We searched and found no Thai instrument mandating an oil containment pit for an outdoor transformer — not in EIT clause 6.4(d), not in ประกาศกระทรวงอุตสาหกรรม 2552, and not in กฎกระทรวงแรงงาน 2555.

And this three-times rule was read from training material reproducing the standard, not from the printed EIT 022001-22 volume. Verify before citing it as binding.

What NFPA 15 actually requires about controlling discharged water

Clauses 4-2.1 to 4-2.3 of the 1996 edition require discharged water to be controlled or contained wherever flammable or combustible liquids are present, and list five acceptable methods: (a) curbing and grading, (b) underground or closed drains, (c) open trenches or ditches, (d) dykes or impounding basins, or (e) a combination.

Two specific requirements belong on the drawing. Closed drains must have traps or another suitable means to prevent flame or burning liquid entering the system. Open trenches must be routed so that they do not place fire fighters, critical equipment and piping, or other property at risk.

The five things the volume must be calculated from

Clause 4-2.4 of the 1996 edition requires the volume to be calculated from the combined flow of:

1. All water spray systems intended to operate simultaneously in the fire area, using the actual discharge rate where it exceeds the design rate.
2. Hose streams and supplementary monitors likely to be used.
3. The largest foreseeable leak or spill of process liquid.
4. Normal process liquid or cooling water drainage entering the drainage system.
5. Rainwater, where local conditions make it appropriate to include it.

The fifth item is the one that decides everything in Thailand. The standard leaves rainfall to the designer, and no Thai page addresses it. An open pit beneath a transformer in the monsoon may be nearly full before any fire starts.

Annex A-4-2.5 sets a practical ceiling: a major fire in a large plant may run 8 hours or more, and where calculating on that basis produces an impractically large basin, 4 hours is generally taken as the practical ceiling. For spills not exceeding 500 gallons (1,893 litres) with good drainage and containment, the expected duration may be as short as 30 minutes to 1 hour.

The stone layer, and the 50 mm that makes it work

An important caution first: NFPA 15 contains no requirement for a stone-filled pit at all. That detail comes from NFPA 850 and from IEEE, not from NFPA 15. If a page claims it as an NFPA 15 requirement, an engineer opening the standard will not find it.

What evidence exists is as follows, and all of it is secondary — we did not open NFPA 850 or IEEE 979 ourselves.

• IEEE PCIC 2022-PCIC-0545 cites IEEE Std 979 clause 8.2 to the effect that when the oil level lies within 1.5 inches of the top surface of the stone layer, flaming combustion occurs; the pit should therefore be designed so that the oil surface sits at least 2 inches (about 50 mm) below the top of the stone.
• The same paper cites NFPA 850 Annex C.6.2, reporting tests on stone-filled oil pits in which oil standing above the stone level burned, but the fire self-extinguished once the oil level fell about two inches below the top of the stone.

The engineering principle is that the stone acts as a flame arrester by keeping the liquid surface below a depth at which vapour and air can sustain flame. That only works if the level is actually maintained below the stone — which is a drainage and capacity question, not a stone question.

Detection and release logic: why transformer protection relays should not trigger the deluge

This is the section that decides whether the system is an asset or a liability, and it contains the clearest engineering disagreement in the subject.

The response times the standard sets, and three numbers people confuse

NFPA 15 sets several time requirements, and they are different numbers.

The operative clause 4-8.3 requires the detection system to actuate the deluge valve without unnecessary delay, and clause 4-1.3 requires the system and water supply to deliver effective spray from all open nozzles without delay. Annex A-4-1.3 adds that achieving this within 30 seconds may require remote starting of the fire pump, and states plainly that fitting a timer will delay operation and adversely affect performance. The 1969 edition already carried the 30-second figure, so the thinking has been settled for over fifty years.

Acceptance testing must also record the elapsed time until water reaches the most remote nozzle. That belongs in every site acceptance test document.

The disagreement that has to be stated plainly: what should trigger release

FIST Volume 3-32 of the US Bureau of Reclamation, November 2016, states directly that current industry practice does not recommend using most transformer protection devices to initiate a fire suppression system. The devices it identifies as unsuitable are:

• transformer differential relays
• sudden or fault pressure relays
• winding temperature sensors
• oil temperature sensors
• low oil level sensors
• Buchholz relays
• bladder failure relays

The reason given is that inadvertent operation of these devices causes inadvertent operation of the fire system, which contaminates transformer surfaces and bushings and washes spilled oil into watercourses.

The devices the document says should be used to isolate the transformer and initiate suppression are heat and/or flame detectors suitably located near or on the transformer, together with manual release by control switch or push button.

The figure that gives this argument weight

The IEEE Power System Relaying Committee report on Sudden Pressure Protection for Transformers (report number 009) records that a large North American utility discontinued tripping from most of its sudden pressure relays after an analysis of operating history found a misoperation rate exceeding 80 per cent.

The causes the report identifies include through faults causing winding movement and pressure waves in the oil, seismic events, starting and stopping of oil cooling pumps, and maintenance activities such as changing nitrogen bottles. Note that none of these involves a fire. If that relay is wired to trigger the deluge, several hundred litres per minute land on a healthy transformer that may still be energized.

If it genuinely cannot be avoided, the risk can still be reduced

The same report lists accepted mitigations: supervising the sudden pressure relay trip logic with an overcurrent relay to reduce operation on high external fault current and in seismic areas; introducing a short deliberate time delay to avoid operation on cooling pump start and stop; disabling the trip or converting it to alarm only during maintenance; and using Form C auxiliary contacts to reduce false trips from contact bounce.

De-energize first, then discharge

FIST 3-32 states that firewater should not be discharged onto an energized transformer and should not be used as a cooling method, and that the transformer should be automatically isolated before water is released. The interlocks the document lists are:

• isolate the transformer from the system
• block the drain and stop pumping oil-contaminated water from the pit to any watercourse
• stop transformer fans and oil pumps that could fan and feed the fire

Those three items are a checklist an existing system owner can apply to their own installation today.

Two-detector confirmation logic

FIST 3-32 states that heat detection is the most reliable means of initiating suppression for a transformer, that techniques worth considering include linear heat sensing cable and infrared detection, and that depending on detector type and design detail it may be appropriate to require two detectors to operate before initiating suppression, to reduce false operation.

That is the engineering reason for cross-zoned or 2-out-of-2 logic in a deluge control panel, and it should be stated explicitly in a TOR together with a cause-and-effect matrix.

The side effect that turns a good system into a bad one

FIST 3-32 records directly that the deluge systems in that agency’s power plants extinguished transformer fires effectively, but that the very large volume of water discharged readily overflowed containment and washed oil into watercourses. The measures it recommends are timed release and installation of high-level detection in the containment structure to shut the suppression system down before overflow.

That is the design answer to the situation in the calculation section above. When enlarging the pit to a 120-minute basis is physically impossible, the answer is controlled discharge with a high-level cut-off — not shutting one’s eyes and letting it overflow.

Comparing three approaches: separation and barriers, water spray deluge, and nitrogen injection

Comparison table

Approach 1: separation and barriers, the cheapest answer if the layout can still change

The figures we verified from freely available primary documents are UFC 3-600-01 Table 4-5, for separation between transformers and other equipment including other transformers, mineral oil case:

Table 4-4 gives separation from transformer to building, varying with the type of wall exposed, mineral oil case:

The distance is measured not from the tank but from the components that will rupture and spray burning oil. UFC also requires no openings in the lower storey wall within 10 ft (3 m) of a transformer, with existing windows filled with brick or concrete block, and projecting eaves of non-combustible material. These are measures that can be taken immediately without buying any equipment.

For NFPA 850, the criterion cited is clause 5.1.4 of the 2020 edition: adjacent oil-insulated transformers containing 500 gallons (1,893 litres) or more should be separated by a 2-hour fire barrier or by the distances in Table 5.1.4.3, with the barrier extending at least 1 ft (0.31 m) above the tank top and conservator, and at least 2 ft (0.61 m) beyond the width of the transformer and its radiators, or to the edge of the containment area, whichever is greater.

That last phrase — or to the edge of the containment area, whichever is greater — is the part most often missed in Thai specifications. A wall dimensioned only to the tank and radiators fails immediately once the oil pit is wider than the transformer, which is the normal case.

Limits of the evidence: we did not open NFPA 850 ourselves. The wording above comes from an article quoting the clause verbatim, and the 25 ft / 50 ft distances are independently corroborated by UFC Table 4-5, which we read directly. Before printing an NFPA 850 clause number into a contract document, verify against the edition you have bought — and note the status change described earlier, from Recommended Practice to Standard in the 2026 edition.

Approach 2: water spray deluge — what it buys back

The IEEE PCIC paper cites IEEE Std 979 clause 8.1 to the effect that containment structures should use a 3-hour fire resistance rating, but that 2 hours is acceptable where an automatic fire suppression system is installed on the transformer. That is the clearest form of the trade: the fire system buys back one hour of fire resistance.

Set against that, it brings the ongoing obligations of NFPA 25 — valve testing, nozzle inspection and pump maintenance — and the containment problem described in the calculation and detection sections above. A deluge system that overflows its pit converts a fire into an oil spill.

Approach 3: nitrogen injection

Nitrogen injection fire prevention systems work by depressurising the tank and injecting nitrogen to stir the oil and suppress combustion. They are proprietary systems designed together with the transformer, they are not listed as an option under NEC 450.27, and we found essentially no Thai-language operating experience to cite. We note the approach exists; we make no claim about its performance.

A checklist for writing a TOR and reviewing transformer fire system proposals

Use this both when writing the specification and when reading the bids. The items a bidder cannot answer are the ones that become variation orders later.

A. Design basis

• State whether the project requires a Fire Protection Design Basis Document (FPDBD). The 2026 edition of NFPA 850 makes this document mandatory. If the project’s design basis cites NFPA 850, the TOR must say who produces the FPDBD and what it covers.
• State one base standard and its edition — for example NFPA 15, edition in force at the date of tender — and say whether NFPA 850, UFC 3-600-01 or insurer guidance is a supporting document. Do not mix multiple bases in one table.
• State both densities: the value on the prism envelope and the value on nonabsorbing ground, and which applies where.
• Define the “ground surface area expected to be affected” numerically and as a figure on the drawing. Do not leave it to the bidder’s interpretation.
• State the required discharge duration in minutes, and state how much hose stream demand is included.
• State which transformers must be assumed to operate simultaneously.

B. Hydraulic calculation

• The prism envelope area must be shown with the dimensions used, not as a bare area figure.
• The nozzle K value must be stated with its unit system, and the formula used must be shown.
• Hydraulic junction balance within 0.03 bar.
• Show the treatment of velocity pressure, or show that it does not exceed the 5 per cent threshold.
• Use the Hazen-Williams equation and state the C value used per pipe type.
• Show the pressure at the most hydraulically remote nozzle and confirm it is not below 1.4 bar.
• Show the actual discharge rate of the proposed nozzle layout, not merely the minimum density figure.

C. Nozzle and pipe layout

• Nozzle spacing not exceeding 3 m unless listed for greater spacing.
• Gaps between components exceeding 305 mm must have separate nozzle protection, indicated on the drawing.
• The conservator, oil pumps and special arrangements must appear in the protection layout.
• Written confirmation that spray does not envelop energized bushings or surge arresters by direct impingement — or attach written permission from both the transformer manufacturer and the owner.
• No piping across the tank top or the control cabinet face.
• Show electrical clearances at the critical positions for the project’s voltage level.
• Surfaces the spray cannot reach must be protected by a method the standard permits. No rundown credit.

D. Equipment and listing

• Nozzles listed by UL or FM Global, per มอก. 2541 เล่ม 8 clause 7.4.1(1).
• Control valve outside the protected area and safe from fire.
• Spare nozzles per the 6 / 12 / 24 criterion.
• State the listing standards for the fire pump, driver and controller.
• State the strainer type, location and cleaning method.

E. Detection and control logic

• State the detector type and mounting position on or near the transformer.
• State the confirmation logic — for example two detectors before release — together with a cause-and-effect matrix.
• State explicitly whether transformer protection devices may initiate release. Our recommendation, following FIST 3-32, is that differential relays, sudden pressure relays, Buchholz relays and temperature or level devices should not. If the owner requires otherwise, the mitigations listed in the detection section should be specified.
• Require the interlock list: isolate the transformer before discharge, block the pit drain, stop fans and oil pumps.
• Require recorded elapsed time to the most remote nozzle at acceptance testing.

F. Containment

• State the pit volume basis and every component counted: spray water, hose streams, oil, rainfall intensity.
• State whether timed release and a high-level cut-off are required.
• State how oil-contaminated water is handled after the event.

Transformers are not under the same law as LPG tanks or fuel depots — do not swap them

Thailand has three sets of requirements that look similar and are frequently confused. Taking figures across from one to another is a common and expensive error.

Transformers. No Thai instrument specifically mandates a fixed fire system for an outdoor installation. The design basis comes from NFPA 15 and NFPA 850 together with the requirements of the owner and the insurer, as set out throughout this page.

LPG storage tanks fall under an entirely separate body of law, with its own notification setting cooling water density, nozzle orifice size, stored water quantity per unit area, and pump type. None of those figures appears on this page, and none of them applies to a transformer.

Fuel depots fall under a third set again, governing foam systems and cooling water for atmospheric storage tanks.

None of the three substitutes for the others, and figures do not transfer between them. If a document quotes an LPG cooling-water density in a transformer specification, or a transformer density in a depot specification, it is citing the wrong instrument.

What SATU Innovative can do on transformer fire system work

Let us be clear about what is what.

What we do. SATU Innovative works on the engineering, detection, pumping and maintenance side. Our services page states that we provide detailed engineering to NFPA standards and Thai law — foam systems, deluge and water spray systems, fire hydrant networks, fire pumps and clean agent systems (IG-100, CO₂, FM-200) for control rooms, substations and data centres — delivering P&ID and general arrangement drawings, hydraulic calculations to NFPA 13 / 15 / 11, equipment specifications and a BOQ with Factor F pricing.

On systems and maintenance we cover sprinkler systems, deluge and water spray systems, standpipes and fire mains, and fire pump sets, with hydraulic calculation, installation and site acceptance testing, plus NFPA 25 maintenance covering fire pumps, sprinkler and deluge systems, foam systems and fire mains.

For pump sets we supply against testing-laboratory listings: pump to UL 448, diesel engine to UL 1247, controller to UL 218 and transfer switch to UL 1008.

On detection, which is the layer that decides whether a deluge system operates at the right moment, SATU Innovative designs, supplies, installs, commissions and maintains linear heat detection using AP Sensing distributed temperature sensing equipment in Thailand — the product page already identifies cable trays and transformers as applications. For outdoor flame detection, SATU Innovative supplies, installs, commissions and maintains fire and gas detection using Consilium equipment in Thailand, including Micropack FDS301 and FDS303 explosion-proof flame detectors certified to FM 3260 and EN 54-10. Both of these are detection solutions we have selected.

On foam concentrate, SATU Innovative supplies SOLBERG fluorine-free foam concentrate in Thailand; the product page identifies VERSAGARD AS-100 3×3 (Foam-Water Sprinkler) for transformer and electrical fires. Our foam position is fluorine-free only.

What we do not claim. We are not the distributor of any brand of deluge valve or water spray nozzle, and we do not claim to be. What we do with that equipment is specify it against the correct listing standards, check bidders’ certification documents, calculate the system so that it works with the equipment selected, and take responsibility for the performance of the whole system at acceptance testing.

If you are writing a TOR for a transformer yard, or you have an existing system and are unsure whether its detection and interlocks are correct, talk to our engineers with the transformer rating, oil volume, voltage level and the yard layout you have.

Frequently asked questions

What is a deluge system, and how does it differ from an ordinary sprinkler?

A deluge system is a type of piping arrangement, not a type of nozzle. All nozzles are open, the pipework holds no water in the normal state, and on detection a control valve opens and every nozzle discharges simultaneously. An ordinary sprinkler holds water in the pipe and each head opens individually when its own heat element operates. For a three-dimensional oil fire that runs and spreads, individual head operation is too slow and too local, which is why transformers use deluge.

What is a deluge valve and where must it be installed?

It is the control valve that holds water back and releases it on a signal from the detection system. มอก. 2541 เล่ม 8 clause 7.4.2(2) requires the control valve to be installed outside the protected area and safe from fire. That is not a detail — a valve that can only be reached across the area that is burning cannot be operated or reset.

Does Thai law require a fire system on an outdoor transformer?

No. We found no Thai instrument that mandates a fixed fire-extinguishing system for an outdoor transformer installation. กฎกระทรวง ฉบับที่ 33 ข้อ 20 applies to transformer rooms in high-rise and extra-large buildings, which is an indoor requirement. What Thai law does provide is a firewater duration floor — 30 minutes under ประกาศกระทรวงอุตสาหกรรม 2552 clause 10 — and a stored reserve of 9,000 to 36,000 litres under กฎกระทรวงแรงงาน 2555. The design basis for an outdoor transformer comes from NFPA 15 and NFPA 850, plus the owner’s and the insurer’s requirements.

What water density is required for a transformer?

Two densities, not one. NFPA 15 requires 10.2 L/min/m² (0.25 gpm/ft²) over the projected area of the rectangular prism envelope enclosing the transformer and its accessories, and 6.1 L/min/m² (0.15 gpm/ft²) over the nonabsorbing ground area expected to be affected. Almost all Thai material cites only the first, which systematically undersizes the system. UFC 3-600-01 sets a higher figure of 0.30 gpm/ft² (12.2 L/min/m²), so 0.25 is a minimum rather than a universal answer.

Should the bushings be sprayed as well?

No — and this is the most counter-intuitive clause in the standard. NFPA 15 clause 7.4.4.4 requires nozzles to be positioned so that spray does not envelop energized bushings or surge arresters by direct impingement. The only exception is where the transformer manufacturer and the owner both permit it, in writing. Firewater contains impurities, and a wetted porcelain bushing surface is a flashover path. A vendor offering dedicated bushing nozzles as standard practice is proposing what the standard prohibits.

How is a transformer oil pit sized? Can the three-times rule be used?

The Thai three-times rule comes from the EIT standard and is a transformer room rule, not an outdoor yard rule. For an outdoor yard, NFPA 15 clause 4-2.4 requires the volume to be calculated from five components: all water spray systems operating simultaneously, hose streams and supplementary monitors likely to be used, the largest foreseeable liquid spill, normal drainage, and rainwater where local conditions make it appropriate. In the worked example on this page, the oil is less than a tenth of the total once hose streams are counted — firewater, not oil, is the largest part of the volume.

Can a Buchholz relay or a sudden pressure relay trigger the deluge?

FIST 3-32 of the US Bureau of Reclamation states that current industry practice does not recommend it. The devices it identifies as unsuitable include differential relays, sudden and fault pressure relays, winding and oil temperature sensors, low oil level sensors, Buchholz relays and bladder failure relays. An IEEE Power System Relaying Committee report records a utility discontinuing tripping from most sudden pressure relays after finding a misoperation rate exceeding 80 per cent, from causes including through faults, seismic events, cooling pump starts and nitrogen bottle changes — none of which involves a fire. Use heat and flame detection plus manual release instead.

How long must the system discharge for?

NFPA 15 clause 7.4.4.2.3 sets a minimum of the design flow rate plus 250 gpm (946 L/min) for hose streams, for not less than 1 hour. But clause 7.4.1 sets the governing principle that the system must perform for the duration of the expected exposure fire, and the annex notes that several hours may be required. UFC 3-600-01 requires 2 hours. มอก. 2541 เล่ม 8 gives 120 minutes for extra hazard group 2, which a transformer yard holding tens of thousands of litres of oil has a strong case to be. One hour is a floor, not a target.

Is it true that NFPA 15 sets a 15-minute minimum?

Not for transformer work. The 15-minute figure sits in chapter 9 of the standard, covering Ultra High-Speed Water Spray Systems for explosion hazards requiring discharge within fractions of a second. It is not the transformer deluge provision. Anyone writing 15 minutes into a transformer specification is citing the wrong chapter.

Can an inert gas or clean agent system be used on an outdoor transformer?

No. There is no building envelope to hold the agent concentration, and these agents give no cooling effect on hot steel surfaces. Several machine-translated Thai pages propose Argonite or Inergen for substations, which takes a control-room system and offers it as the answer to an outdoor yard problem. Clean agent systems have a real place in indoor transformer and electrical rooms; they are not an answer outdoors.