How to electrical test (PAT test) large transformers?

  I presume these are not really used like portable appliances and as you note the normal PA tester with a 13A socket on it won't quite cut it ;-)

The first thing is to decide what needs testing, what failure modes you hope to pre-empt and then what the test limits should be.

So for class 1 enclosures, the normal earth/ CPC connection can be verified but it needs to be good enough to operate the circuit fuse or whatever is in the supply - no good having an earth connection that breaks at a lower current than the supply fuse ;-) to that end certainly at the 90KVA end, I'd expect to see earthing of the same cross-section as the live cores, probably 25mm2 assuming that is a 3 phase unit, or rather more if its single phase. testing that with a meter is not really that easy or desirable - and if you do it will require a 4 terminal instrument. A quick buzz test with the lowest ohms range your normal meter can do,  and a thorough visual inspection may be more appropriate. Finding 'high' resistances on the scale of tens of milliohms is usually fraught with difficulty.

Insulation primary to core/frame and from secondary to both core/frame and primary should be confirmed (except autotransformers and variacs, where primary and secondary are the same coil). Expect several megohms, but don't test at more than 500V DC without checking with the makers. It should do a lot better than that, but how much better is not clear.
If you can measure the magnetising current without a load connected, and I realise that may not be so easy, where loads are wired in, as any increase in this between successive tests indicates a shorted turn or deterioration of the insulation between laminations.
M.

Very useful, thanks.  How could one measure the magnetising current?  Never heard of that before

To measure magnetising current, connect the transformer to a supply of the correct voltage and frequency, and disconnect all loads from the secondary side. Measure the input current. Basic testing is usually done at nominal voltage and frequency, more detailed testing may be done at the upper limit of permissible input voltage such as 253 volts (110% of 230 volts) and at the lower limit of frequency, such as 49.5 cycles, the lower limit of UK grid frequency.

Magnetising current increases at high input voltages and low input frequency. Excessive magnetising current may indicate an internal fault.

basically, without a secondary load the only current that flows is that required to magnetise the core back and forth on alternate cycles - the  no-load primary current. this should be a lot less than the on-load current, or some fault in the transformer is acting as an extra load.

It is also referred to as the primary inductance. The similar sounding but complementary quantity, the "leakage" inductance, is the input inductance when the secondary is shorted rather than open circuit, but for most transformers you want to keep operational for more than a few cycles, there are very good reasons for calculating that from voltage droop tests, not trying to measure it directly !

Mike.

Leakage inductance may be measured by short circuiting one winding of the transformer, and applying a much reduced voltage to the other winding. Adjust this voltage until normal load current passes.

I have done this, years ago, a few hundred volts was applied to the 11kv side of a distribution transformer, whilst the LV side was short circuited.

Leakage inductance may be measured by short circuiting one winding of the transformer, and applying a much reduced voltage to the other winding. Adjust this voltage until normal load current passes.

I have done this, years ago, a few hundred volts was applied to the 11kv side of a distribution transformer, whilst the LV side was short circuited.

This is very true - thinking back I too have seen this sort of thing done, though only as a customer witness, with something like 230 or 110V on the 7/11kV side of things. Rather like running a current transformer with no load one has to know exactly what one is expecting and in my defence I was being a bit quick as I wanted to keep the explanation simple, but just mention it in case the OP had seen both inductance terms somewhere and to point out that there are 2 very different parameters not to be confused, and different tests.

A fall in magnetising inductance - an increase in no load primary current is a good indicator of trouble cooking.
M.