I'm not a relay engineer. I'm the person who commissions transformer protection systems and then gets called when something trips. For eight years, that was my whole job. In that time, I personally made—and documented—six significant mistakes that added up to roughly $180,000 in wasted budget and downtime. I now keep a checklist. In the past 18 months, that checklist has caught 47 potential errors before they became problems.
The most important lesson? Most transformer protection 'failures' are not the transformer's fault.
The Surface Problem: The GE Multilin 850 Kept Saying 87T
At 2:13 AM, a 10 MVA GE transformer tripped. The alarm was clear: transformer differential, 87T, on a GE Multilin 850 relay. The front screen showed differential current at 3.2 per unit. We reset the relay, re-energized, and 40 minutes later it tripped again.
Six hours later, the oil sample came back clean. No dissolved gas. No winding movement. No internal fault. The transformer was fine.
The relay had not failed. The relay was telling the truth, based on the information it was receiving. The problem was that the information was wrong.
That night cost $17,400 in overtime, testing, and lost production (not counting the hit to my pride). I started calling it 'the night the relay wasn't wrong.'
The Deeper Reason: The Relay Only Knows What the CTs Tell It
Here's the thing: differential protection is simple arithmetic. The relay compares current entering the primary with current leaving the secondary, adjusts for the transformer's ratio and vector group, and, if the remaining difference is too big, opens the breaker.
Every GE Multilin 850 relay transformer protection scheme depends on the same assumption: the current entering the relay represents reality. It does not know whether the transformer is actually healthy. It does not care how expensive the transformer is. It only knows what the current transformers (CTs) and wiring deliver to its terminals.
When the relay says '87T,' the correct question is not 'Is the relay broken?' It's 'What input convinced the relay that there was an internal fault?'
Mistake #1: I blamed the relay first. The cause was a reversed CT.
My most expensive mistake was on a 25 MVA GE transformer with a Multilin 850 and an 87T scheme. We were running a load test, and at 30% load the differential current jumped from 0.2 per unit to 4.6 per unit. I checked the CT ratios. I rechecked phase rotation. I called GE support and followed every step. Still 4.6 per unit.
I went back and forth for two weeks between 'it must be the relay' and 'it must be the wiring.' On paper, the wiring was correct. The old drawing had a mirrored CT polarity mark. One wire. That's what the problem was—one reversed neutral CT lead.
Polarity. That's it.
The relay performed exactly as designed. It saw current leaving where current should have been entering, and it calculated a massive differential. If I had replaced the relay, the problem would have followed the wiring.
I had already ordered a replacement relay and paid for emergency delivery before I found the reversed wire. The unused relay, test jumpers, and overnight shipping invoice came to $8,200. The wire that fixed it cost nothing.
Mistake #2: The vector group setting was wrong.
Another time, the relay was tripping on energization. Not during a fault. During magnetizing inrush. The 850 has second-harmonic restraint to prevent that, but the setting was disabled.
Why? A few weeks earlier, someone had set the relay logic to 'simplified' for a temporary test. They disabled harmonic restraint to make the relay respond faster. They forgot to turn it back on. The relay saw the inrush current, decided it was a fault, and opened. The transformer was fine.
This gets into relay setting territory, which is not my expertise. I'm not a protection engineer, so I can't walk you through a full harmonic restraint stability study. What I can tell you as the person commissioning the equipment is this: before you energize a GE transformer, verify the setting group is the one you actually want. A temporary test setting can become your permanent trip log.
Mistake #3: CT saturation was silently making the 87T element unstable.
There were also the cases where everything in the wiring was right, and the relay still tripped on through-faults. The culprit was CT saturation. During an external fault, a CT can saturate, distort the current waveform, and create a false differential current. The relay can't tell whether the distortion came from a transformer fault or a CT that's too small for its burden.
This is the hardest one to find, because the relay doesn't give you a 'CT saturation' alarm. It just gives you a trip. On old protection-class CTs, this is a real risk. If you're using a modern GE Multilin 850 relay with older CTs, don't assume the relay can compensate for CT limitations. It can't.
If you ever google 'how to access NVIDIA Control Panel', you know the feeling of looking for the obvious and finding nothing. The 850's menu has the same learning curve. But the relay's menu isn't the issue. The issue is what's upstream of the relay.
The Real Cost of a Wrong 87T Trip
One false 87T trip is not just 'reset and watch.' It brings an oil test. It brings a technician call. In 2023, one site lost a 12-hour production window and paid $6,100 for the service call and testing. The transformer was healthy. The settings were wrong.
Small plants and small utilities get the worst of this. They often don't have a full-time protection engineer, so they get told 'the relay is bad' or 'the transformer is bad.' I've seen vendors replace a perfectly good relay because nobody checked the event record. That's not how you build trust.
Take it from someone who made those mistakes: the small orders were where I learned, and the vendors who took me seriously then are the ones I still call first. A 1.5 MVA transformer in a factory is just as critical to that factory as a 100 MVA unit is to a utility.
What I Check Before I Blame Any Transformer Relay
In Q1 2024, after the third false trip in two months, I built my current pre-energization checklist. I based it on my own screw-ups and on two sources: IEEE C37.91, the guide for protecting power transformers, and GE's 850 instruction manual. Neither one says to trust the inputs. They say to verify them.
- Pull the event record before you reset it. The 850 stores waveform captures. Look at all three phase currents. Did the differential current appear before the trip? Did one phase, or all three? This tells you more than the display ever will.
- Trace CT polarity physically. Don't trust the old drawing. I've seen mirrored drawings, swapped polarity marks, and neutral CTs wired backwards. This was my mistake #1.
- Confirm the relay's vector group and phase compensation match the transformer nameplate. On a GE transformer, the vector group is on the nameplate. The 850 setting must match it. A mismatch looks just like an internal fault.
- Verify CT class and burden. If you can't do the saturation calculation, ask someone who can. That's the part I'd involve a protection engineer for.
- Check harmonic restraint and temporary settings. If anyone touched the relay before the trip, start there. The second-harmonic restraint exists for a reason.
And yes, check the surge protection too.
What does a surge protector look like in a transformer cabinet?
It does not look like the Home Depot surge protector you plug a TV into. If someone says 'we added surge protection' and shows you a power strip, that's not the same thing. In a substation, a surge arrester for the transformer terminals is usually a porcelain or silicone-housed block mounted on the structure or directly on the transformer bushing. For control circuits, the surge protective device is a DIN-rail module inside the relay cabinet, often marked with UL 1449. It protects the relay's power supply and CT circuits from voltage transients. The Home Depot version doesn't belong in that cabinet.
Use a UL 1449-listed Type 1 or Type 2 SPD with the correct voltage rating for the control circuit. It should be a component designed for the panel, not an extension-cord accessory.
Bottom Line
The GE Multilin 850 is a solid relay for transformer protection. It has settings, event records, and protection elements that are genuinely useful. But the relay is not magic. If the CT wiring, polarity, vector group setting, or CT capability is wrong, the 87T element will do exactly what it was designed to do: trip on a difference that isn't real.
The next time your relay says 87T, don't ask 'What's wrong with the relay?' Ask 'What information did the relay see?' Then work backward.
It took me $180,000 and six documented mistakes to learn that. Start with the checklist above. It's much cheaper.
Leave a Reply