You've eliminated several possibilities. Troubleshooting the intermittent ground fault, though, might come down to - "When you have eliminated all which is impossible, then whatever remains, however improbable, must be the truth.
" ~ Arthur Conan Doyle, The Case-Book of Sherlock Holmes. Remember, like Sherlock Holmes, by being a technician looking for a problem, you are a detective.
Also, because you have established that you can duplicate the problem, you have some control over the detection process. That is important because you don't have to wait random times between trying each possible cause of the ground fault light turning on.
Make the assumption that there is a problem: the ground fault light sometimes turns on. This is the only assumption you can make about the cause of the problem, until you have it fixed.
- There could be a single intermittent cause of trouble that turns on the ground fault light
- There could be more than one intermittent cause of trouble, and each of them, by themselves, could turn on the ground faut light
- There could be multiple intermittent causes of trouble, but two or more have to be occurring at once in order to turn on the ground fault light
There are several possible places where the problems could be coming from.
- The batteries (with hairline cracks) could be a source of the problem(s)
- The wire dressing (what wires inside the control panel are next to each other) could be a source of the problem(s)
- The wires coming into the control box could be a source of the problems
- The control panel's circuit board could be a source of the problem(s)
Easy Stuff First
When eliminating the impossible, start with the easy stuff. Make sure the problem isn't something easy to find. Start with the batteries. Making sure the batteries aren't leaking some current to ground is really easy.
Use some cardboard and lift the batteries off the panel's back-box. Make sure the batteries, including the battery wires, aren't touching the box, or any other wires in the panel. That makes sure the batteries are not bringing the ground fault into the panel's circuit board. Keep the cardboard inside the panel. It won't hurt.
Inside the control panel, make sure that the utility power wires (power unlimited wires carrying the AC power) aren't close to any of the rest of wires (power limited wires carrying the inputs and outputs to the circuit board inside the panel). If possible, even the battery wires should be separated, at least a little bit and not bundled with any other wires.
Then, making AC voltage measurements relative to ground, use your AC voltmeter to make sure that each of the wires, except for the power wires, don't have any AC voltages on them. To make these measurements, connect the black lead of your AC voltmeter to a bare screw in the panel's back box, and use the red lead to check, one at a time, the screw terminals for each wire landing on the printed circuit board.
Normally, none of the fire alarm's wires, except for the power wires, have any AC voltages relative to the cabinet ground.
The relay output screw terminals, common, normally open, normally closed, and any telephone wire screw terminals, are not connected directly to any of the actual building wide, Fire Detection and Alarm System. Ignore, for now, any AC voltages on those screw terminals.
When doing the measurements, write down the AC voltage on any screw terminal that has even 0.1 AC volts on it. This is a fire alarm system and there shouldn't be any AC voltage (relative to ground) on the system.
You are changing something or another when you turn on and off the utility power. Something is changing slightly; the problem is always there, but turning on and off the utility power causes the ground fault light to turn off and on.
I've worked on several panels that, when I'm on site looking inside the panel, trying to figure out what's wrong, the system goes normal. It stays normal the entire time I'm looking at it. Then, 10 minutes to half an hour after closing the panel and leaving, the system starts having trouble again.
What I was actually doing, though, was changing the temperature inside the control panel. On arrival, by opening up door to the panel, I lowered the temperature a few degrees on the circuit board, and the panel became stable and normal. When leaving, by closing the door to the panel, the temperature on the circuit board started to climb a few degrees, and the panel started to have troubles again.
The problem on these panels turned out to be a marginal thermal problem. Slightly lower air temperature, and the panel is normal; slightly higher air temperature, and the panel has trouble.
In these cases, because opening and closing the cabinet door would cause the problem to come and go, replacing the panel's circuit board fixed the problem. It's an easy but expensive problem to fix, but it's a hard problem to find in the first place.
Power Supply and Ground Fault
All fire alarm panels have power supplies built into them. The ground fault detection circuitry is usually part of the power supply, so the ground fault circuitry is close, physically, to the power supply. When the utility power is turned on, the heat of the power supply affects the ground fault detection circuitry, at least a little.
With your problems, when the ground fault circuitry is warm (the utility power is turned on), the circuitry might work; when the ground fault circuity is cool (the utility power is turned off), the circuitry might fail, and a failure turns on the ground fault light.
This is one of those improbable faults that can only be determined by eliminating everything else, leaving the improbable coolness from the power supply as the only possible one.
Take my word for it, pointing a finger to that kind of problem isn't an easy decision. Make sure you've eliminated all of the other possibilities first.
Check for Induced AC Voltages
I have had the ground fault light turning on and off, seemingly randomly. The cause was the fire alarm wires occupied the same conduits as the power circuitry for the air handlers. Even though the wires were properly insulated, the air handler wires would magnetically induce an AC voltage into the fire alarm wires. Like how any transformer works, AC power was transferred magnetically from one set of wires to the other set.
The original fire alarm system was old, and ran entirely off utility power. At that time, running the fire alarm wires in the conduit wasn't a problem, either technically or legally. When the fire alarm system was upgraded, the fire alarm system no longer used utility power, but the wires for the fire alarm system were never removed from the conduit.
By leaving the wires in the conduit, the AC current from the air handlers induced a low current-voltage into the fire alarm SLC.
This was found by seeing a small AC voltage, measured against ground, on the Signaling Line Circuit (SLC) input to the panel. The voltage was about 0.6 volts AC while the SLC was connected to the panel. However, when both wires were disconnected, the induced voltage on the wires themselves jumped up to 12 volts AC.
The solution was to route the SLC wires around the affecting conduit, which eliminated the magnetic AC coupling between wires.
Remember that earlier in this email there was the suggestion to mark down the screw terminals that showed any AC voltage relative to ground? Now, disconnect the wires that have shown voltage, one pair at a time, and measure the AC voltage relative to ground on the wire. If there's a lot of voltage, relative to ground, your problem may be coming in there.
This should be checked before concluding that the panel's expensive circuit board needs replacement.
Make sure you've eliminated all the normal possibilities, the ones that you can easily verify. Then you can consider improbable things like a marginal thermal problem or an induced AC problem; problems that are hard to directly verify.
By eliminating all of the probable causes of the ground fault, the highly improbable causes of the ground fault seem to become more plausible.