A client operates a 22/0.4 kV substation from 1986 in an industrial estate. Oil-filled distribution transformer 1,000 kVA, manufacturer BEZ Bratislava. A year before the planned switchgear modernisation, the energy engineer brings three tests: BDV (Breakdown Voltage Test) shows 38 kV/2.5 mm instead of the required 50+ kV. DGA (Dissolved Gas Analysis) shows 380 ppm acetylene. Thermography shows a hot spot at the HV winding inlet. Question: repair for €15–25k or replace for €40–60k? Here's the decision grid that walks every piece of diagnostics before the PO is signed.
Diagnostic tests — what they actually say
The prerequisite for any decision is a complete diagnostic package. Without these numbers the decision is guesswork.
BDV — Breakdown Voltage Test (IEC 60156)
Oil dielectric strength test. An oil sample is placed between two electrodes 2.5 mm apart and voltage is ramped up until breakdown. The value = the voltage level at breakdown.
- **New oil**: ≥ 70 kV/2.5 mm.
- **Acceptable**: 50–70 kV.
- **Questionable**: 40–50 kV. No further degradation tolerated, plan filtration.
- **Unacceptable (critical)**: < 40 kV. The oil is wet, contaminated or oxidised. Breakdown possible during operation at normal voltage.
Key point: BDV measures mainly moisture. Dry oil (< 10 ppm water) typically gives 60+ kV. Wet oil (> 30 ppm water) can drop below 40 kV. **Oil filtration solves BDV for €3–5k on 2,000 l of oil**. But BDV doesn't cover thermal degradation — DGA does.
DGA — Dissolved Gas Analysis (IEC 60567)
Test of dissolved gases in oil. Under thermal and electrical degradation processes the oil in the transformer generates specific gases:
- **H₂ (hydrogen)**: low-temperature corona or partial discharge.
- **CH₄ (methane)**: thermal degradation of the oil at 150–300 °C.
- **C₂H₆ (ethane)**: thermal degradation of the oil at 250–450 °C.
- **C₂H₄ (ethylene)**: thermal degradation at 500–700 °C (overheating).
- **C₂H₂ (acetylene)**: high-energy arc, > 700 °C.
- **CO, CO₂**: degradation of the paper insulation (cellulose).
**IEEE C57.104 limits** for a 22 kV unit in normal regime:
| Gas | Normal (ppm) | Caution (ppm) | Critical (ppm) | |-----|--------------|---------------|----------------| | H₂ | < 100 | 100–700 | > 700 | | CH₄ | < 120 | 120–400 | > 400 | | C₂H₆ | < 65 | 65–100 | > 100 | | C₂H₄ | < 50 | 50–100 | > 100 | | **C₂H₂** | **< 1** | **1–9** | **> 9** | | CO | < 350 | 350–570 | > 570 | | CO₂ | < 2,500 | 2,500–4,000 | > 4,000 |
**Acetylene is the red alarm.** > 9 ppm means an active arc inside the transformer. **380 ppm**, as in our opening case, is an extremely serious state — likely persistent discharges on the winding insulation that will eventually become a breakdown.
Thermography + ZF / DC winding resistance
- **Thermography**: measured under load. Hot spots > 90 °C in summer at 75% load indicate a local pre-heat (bad joint, winding inlet, periodic dust clumping on the cooler).
- **DC winding resistance**: difference between phases > 2% indicates a loose or damaged joint.
FRA — Frequency Response Analysis
Test for detecting winding deformation (after a short-circuit event, transport, or age). A frequency spectrum 20 Hz – 2 MHz is applied, the transfer function is measured. Compared against a baseline from the factory (often unavailable for old units) or between phases. **Practical use**: if the winding has experienced a short-circuit or strong transient, FRA exposes micro-deformation before classical tests.
Decision tree — when retrofit, when replacement
Retrofit makes sense when
1. **BDV < 40 kV, other parameters OK.** The oil is wet or contaminated, but iron + winding are fine. Solution: **on-site oil filtration + dehydration** (vacuum filtration with activated carbon, continuous 48–72 hours for 2,000 l of oil). Price: **€4–7k** including rental of a mobile filtration unit (Hydroflon CJC, Pall PMC). 2. **DGA mildly elevated (CH₄, C₂H₆, no acetylene), CO/CO₂ within norm.** Indication of mild past overload, no active fault. Solution: **oil regeneration** (Fuller's earth treatment or Sea-Maraviv process), DGA again at 30/60/180 days. Price: **€8–15k.** 3. **Localised thermal hot spot identified by thermography** — typically the HV winding inlet, bad bushing contact. Solution: **bushing repair + joint retorque** + thermal remeasurement. Price: **€2–4k** + a mandatory 2–3 day outage. 4. **Tank tightness poor, no other problem.** Oil leak from the bottom or joints. Solution: **reseal**, clean and top up. Price: **€3–6k.** 5. **Age < 25 years and winding OK per FRA.** Investment in retrofit makes sense — 15–25 years of remaining life.
Replacement makes sense when
1. **Acetylene > 9 ppm on repeated measurement** (3 months, 6 months). An active arc means permanently damaged insulation that oil treatment won't repair. **Premature failure likely in 6–24 months.** Retrofit won't change DGA, it only postpones failure. For a 1,000 kVA unit, the cost of a production stop from transformer failure is typically **€50–200k** (12–72 hours of unplanned outage + service intervention). A planned replacement is better. 2. **CO/CO₂ high relative to C-hydrocarbons.** Indicator of cellulose insulation (paper around the winding) degradation. The paper can't be replaced without complete teardown — uneconomic for units < 5,000 kVA. **Remaining paper life < 5 years under intensive operation.** 3. **Age > 35 years + several borderline parameters.** Even if no single parameter is critical, the cumulation (BDV 42 kV, mildly elevated DGA, mild hot spots, winding resistance variation 1.5%) means you stand on ten fuses. One fault triggers a cascade. 4. **Loss factor at the margin.** The magnetic characteristic of older transformers (1980s) is 25–35% worse than modern ones (CRGO + better design or amorphous cores). **For a 1,000 kVA unit at 60% load, annual losses on an oil-CRGO transformer are ~3,800 kWh, on a modern amorphous one ~2,100 kWh.** At €0.18/kWh the difference is **€300/year**, which over a 15-year life = **€4,500** saving — an argument for replacement even when the old one still works.
Borderline cases (for consultation)
- BDV 38–45 kV + DGA mildly elevated (no acetylene) + age 20–30 years: filtration + regeneration + 6-month monitoring. If DGA rises, plan replacement within 24 months. If it stabilises, another 5–10 years of operation possible.
- Large unit (> 2,500 kVA) with a serious problem: unlike a small substation, a **major overhaul** (re-winding, re-coring) in specialist workshops (Trafostav, ABB Service, KONČAR) makes sense. Cost **30–60% of new** + 8–16 weeks of outage.
New unit — CRGO vs amorphous core
When deciding to replace you'll meet two main magnetic core technologies:
CRGO (Cold-Rolled Grain-Oriented) silicon steel
Standard technology, mass-produced. Dominant makers: ABB, Schneider Electric, Siemens, KONČAR, BEZ Bratislava. Lamina thickness **0.23–0.30 mm**, surface insulating varnish.
- **No-load loss (P₀)** for 1,000 kVA: ~**1.1–1.4 kW.**
- **Load loss (P_K)** for 1,000 kVA at rated load: ~**11–13 kW.**
- **Price of a 1,000 kVA oil-filled CRGO unit**: **€18–28k.**
Amorphous core (Amorphous Metal, Metglas / Hitachi 2605SA1)
More advanced technology, lamina thickness **0.025 mm** (10× thinner than CRGO), amorphous atomic structure reduces hysteresis losses.
- **No-load loss (P₀)** for 1,000 kVA: ~**0.28–0.42 kW** (3–4× lower than CRGO).
- **Load loss (P_K)**: comparable to CRGO (~10–12 kW).
- **Price of a 1,000 kVA amorphous unit**: **€26–38k** (40–50% more than CRGO).
- **Height and weight**: ~15–20% taller than CRGO due to a larger core (amorphous material has lower B_max).
Decision arithmetic: at a load factor of 40–60% and electricity at **€0.16–0.22/kWh**, the amorphous unit's payback through low no-load losses (the unit runs 8,760 h/year connected, even without load):
- (1.2 – 0.35) kW × 8,760 h × €0.18 = **€1,340/year saving.**
- Price difference €10k / €1,340 = **7.5 years.**
Over a 25-year transformer life amorphous **returns ~€33,000 in electricity** — pays for itself 3.3 times. **For units with low load factor (backup transformers, lightly loaded industrial) amorphous is almost always economically advantageous.**
For high-load units (load factor > 80%, like utility distribution transformers) the difference is smaller and CRGO can be on par.
Oil vs dry-type (cast resin)
The second choice on replacement: liquid-cooled or dry transformer.
Oil-filled (Mineral Oil / FR3 Natural Ester / Silicon Oil)
- **Price**: lower (by 15–25%).
- **Noise**: quieter.
- **Efficiency**: comparable or slightly higher.
- **Risks**: oil basin, fire risk (mineral oil = class K3, FR3 ester = K2–K3, silicon = K2). Requires a **drip tray** with capacity 100% of oil + 10% reserve.
- **Suitable for**: outdoor installations, separated transformer rooms with drip provision, sites where full fire separation is solved.
Dry-type cast resin (typically winding cast in epoxy)
- **Price**: higher (20–35% more expensive).
- **Efficiency**: 1–3% lower than oil-filled (higher no-load losses).
- **Noise**: louder (core vibrations transmitted directly to air).
- **Risks**: no oil, no fire hazard (epoxy = class F1). Suitable for **indoor installations in buildings** (industrial production, data centres, hospitals, ships).
- **Suitable for**: environments where oil is unacceptable — IT data centres, basements, warehouses with poor ventilation.
On a switch from oil-filled to dry-type budget **20–30% higher price** + **additional cooling** (forced ventilation, sometimes AC) + **higher noise profile** that may need acoustic damping if the substation is near workplaces.
Real price ranges — final comparison for a 630 kVA distribution unit
| Variant | One-off cost | Outage | Life after | Total cost of ownership (15 years) | |---------|--------------|--------|------------|-------------------------------------| | Filtration + monitoring | €4–7k | 2 days | 5–8 yr | €30–45k (incl. second retrofit in 5 yr) | | Oil regeneration + tank repair | €8–15k | 3–5 days | 8–12 yr | €25–35k | | Major overhaul (re-winding) | €18–30k | 8–12 weeks | 15–20 yr | €30–40k | | New CRGO oil 630 kVA | €15–25k | 1–2 days | 25–30 yr | €18–30k | | New amorphous oil 630 kVA | €22–35k | 1–2 days | 25–30 yr | €18–25k (with electricity savings) | | New dry-type 630 kVA | €28–45k | 1–2 days | 25–30 yr | €30–50k |
For 630 kVA the economics of repair drops radically — **a new unit costs only 2–3× more than basic filtration + monitoring, but gives 25 years of peace against 5–8 years.** At 1,000–2,500 kVA the economics tilt even more toward replacement.
A different story applies to **large units 5,000+ kVA**, where new means €120–300k and a specialist repair (re-winding in KONČAR/ABB Service workshops) at 30–50% of the price is an attractive option.
Tipping point — when to stop pouring money into the old
Stop-loss criteria for retrofit (combination of 2+ of the following):
1. Age > 30 years + acetylene detected in repeated DGA. 2. Third repair in the last decade (filtration + tank + bushing). 3. CO/CO₂ rising year-on-year by > 15% with no load change (paper ageing). 4. Hot spots with no identifiable cause (winding deformation, internal contact). 5. Remaining paper insulation < 50% of original DP (Degree of Polymerisation, measured by an accredited lab, ~€400–800).
At this point stop planning retrofit and invest in a new unit. Every further repair is by definition expensive **and has diminishing return**, because remaining life shrinks.
---
*We developed this decision framework for clients in industry where substations often "get lost" in maintenance planning and slide into emergency mode. If you have a unit older than 25 years and the last DGA test was more than 2 years ago, the first consultation (90 minutes) walks through testing strategy, current parameters and reopens the retrofit-vs-replacement decision before a failure forces it on you.*