I'm working on an installation of luminaires that have electronic power supplies that have a large inrush current at turn-on.  A typical string of lights on a 20A branch circuit (277V) would produce as much as 300A rms for 8.3ms (1/2 AC cycle).

Is a high-inrush type of CB (circuit breaker) the best solution (assuming I can't change the luminaire power supply, etc.)?

What do other people do when they run into these situations?

I'm seeing more and more solid state lighting and some have low inrush and some have high inrush - it seems that this must be getting to be a more common problem as lighting migrates to electronic drivers.

Thanks for any thoughts.

Be aware, then used CB has to match with (should not exceed the overload capability of) the used wiring. As i expect the CB and wiring were specified together matching each other (i doubt the wiring would be thicker then required for original configuration), most likely the wiring would be overloaded by this inrush current too, what might lead to fire.

The only thing you might do is to check the wiring rating and if you find it higher then used breaker, use the CB rating corresponding to used wiring. Be aware, then only the rating of the "narrowest" spot count, even if it supply only one fixture, as in case of fault in such fixture all current would flow here.
If this does not lead to eliminate the inrush tripping, you should upgrade your wiring installation, as the wiring would be overloaded.

Other possibility would be add delay relay to part of the installation, so the inrush would not be so concentrated in time.

If you change the response type for the higher inrush current, you might increase the failure mode thermal load on wiring above the safe level.
Each fusing device (fuse, CB) has it's response curve - the shut OFF time vs the current.
Similar curve result, if you analyze the necessary protection response time as function of current in order to keep wiring safe in such overload event. Obviously, if you want to protect such wiring, the response curve of the CB should be always below the the curve of wiring.
And it "happen" (OK, it is intentionally so), then the "standard" CB's (in Europe marked as "B" characteristic, in US would be similar system) and fuses are designed so, their characteristic match the requirement of most wiring designs rated for the same maximum load.

While "slower acting" or "high inrush current" breakers are designed to provide protection for higher inertia devices (like transformers, motors, etc...) against overload enough to overheat the motor, but not enough tripping the "standard" CB. Therefore special "slower acting" CB's are produced, what have lower current threshold when loaded for longer time.
Unfortunately CB and fuse marking is done based on maximum steady current, that these devices do not trip, what does not mean then any "15A" CB would be able to protect "15A" wiring.

So if you use the CB of delayed action or higher inrush current type, it's characteristics would exceed the requirement of wiring for current faults in such level, yielding to it's overload.

Easiest is to use a delay/motor rated breaker if you can. A pity the manufacturer did not add inrush limiting to the ballasts, but you can get the inrush limiter NTC thermistors from RS ( in Europe, but they probably have distributors in the USA) that you can place in series with the supply to each individual ballast, and these will be very good at limiting inrush current, at the expense of running hot during operation.

I did this to reduce breaker size, as I did not want to have a 40A breaker for a 15A load just because of very low supply impedance ( 40m of 25mm cable to 1MVA transformer) and I did not want to place a line reactor in the 3 phase supply, too expensive and way too big, the 3 NTC limiters run warm, but stopped nuisance tripping totally on the packaging line. They were cheap, although they need to be mounted to screw terminals in free air to dissipate heat, as they can run up to 200C during fault conditions, and at least fail open circuit by cracking into pieces.

I'm not sure what options you have available to you but...
If you have the ability to do so, you maybe able to use a relay control to divide the lights into different circuits.

Another option that we used in a low voltage application was a sequencer.  It was used on an ambulance to turn all the outside flashers on without hitting the alternator with a large load all at once.  When the master switch was turned on, the sequencer turned on small groups of lights with 1 second delays.  Like i said, this was a low voltage application, but I would think something that works like that would be available for high voltage systems as well.

Consult the manufacturer or system designer.  277v is usually not used for motor loads so I'm not sure if they even have motor rated breakers. There are many things that are technically feasible, but commercial building lighting is not something to play around with and whatever you do should comply with whatever applicable codes.  

You maybe able to have the lighting control panel setup so that they don't all come on all at once.  

In one of the big SMPS I took apart,  it used a power resistor in series, which was shunted with a triac once the main capacitors charged up.  Newer Sylvania ballasts as well as some Lutron ballasts also use inrush limiting control (to reduce damage on relay and switch contacts)

Is it causing enough inrush to interfere with 480v loads within the facility?  
A line reactor will certain reduce inrush current by cushioning di/dt rate, but then, I don't know of specific UL rated devices that can be implemented in a NEC compliant way.

@Luminaire: I would be very careful with inductors: These can form LC circuits with ballast tank capacitors, what might cause (in worst case - when started just before the voltage reach it's peak value) factor of 2 overvoltage on the internal DC bus, killing semiconductors behind.
I think there are three options, what would for sure comply with the code, both already mentioned:
1) Using ballast with internally controlled and limited inrush current - does not need any wiring modification
2) Splitting into multiple parts and use sequencer. Relay cascade using power relais with mains voltage coil would be enough - coil connected to "previous" section, connecting to it the "following" section. The ordinary power relay delay (5..20ms) would be enough to make working cascade effect. This require slight wiring modification (adding relay boxes).
3) Upgrading wiring to higher current rating so the use of higher current CB would be possible, or split it to more branches with separate CB's. This would require extensive upgrade work, but it's the simplest, when the installation reconstruction is considered anyway (e.g. due to end of present installation life).


Other technical option would be using central RESISTIVE (to avoid the overshoot mentioned above) turn ON surge limiter, but i'm not sure, if there is any such device certified and available.