At the simplest, all you need is an AC voltage source, some LC components representing the transformer's small-signal parameters, and R+C representing the load.
Core saturation isn't important as that is easily checked separately, but there's no harm in using it in a simulation. Maybe the results generate a little slower (it adds nodes to the matrix, if nothing else).
The inverter square wave isn't even important, because, even if a pulsed waveform makes it to the output, the fundamental frequency will deliver most of the power, and all harmonics will be some percentage of that (typically under 10%). There are some cases where this isn't true, and there are cases where square pulses are simpler anyway (square-pulse type SMPS analysis), but this is an assumption we can validate.
Note that a simple RC load won't reflect the whole situation, because for short time scales, most plasma filaments, present at a given instant in time, remain conductive; therefore there is less resistance at high frequencies. But there is also less reactance at high frequencies. I don't know what the actual balance is, but I would assume just for starters that the effect is constant Q. Thus, we emphasize what the resistance is: an effective series resistance (ESR), not a real physical resistor in the system. (Because, again, it's not a resistor at all, it's myriad very small-value resistors incessantly making and breaking connections to incremental capacitances; it's a time-averaged equivalent of the very intricate microscopic physics going on.)
At least, I'm assuming (or, have been) that you want a corona discharge, consistent with the first request: a plasma tweeter.
Plasma tweeters do NOT use continuous point-to-point arc discharges. Arcs have smaller volume, and, I think something about the acoustics that makes them poorer besides? Certainly, the convection of a fat thermal arc will produce subtle but persistent doppler distortion, on top of increased noise (the roiling boiling sound as the arc wanders and convects); corona discharges still suffer from this, which is why -- besides the ozone and NOx -- they're just not very good at this, but, they are the best case among related mechanisms, so that's the way it has been done.
Apparently an arc doesn't look like a short circuit, otherwise the breaker would cut out on overcurrent:
Right. Several posters here have been less than helpful, posting factoids without relating them to actual in-circuit parameters.
A welding arc has low resistance because it is short length; there is still a built-in potential due to the voltage drop near the electrodes (typically 10-20V depending on gas, electrode, pressure, etc.), and there is a quasi-ohmic drop through the plasma itself. Arcs don't simply have one voltage and that's it. This is highly obvious when you measure the voltage of a welding arc, while moving it around; it grows with distance, which is why stick/TIG power supplies use a relatively high open-circuit voltage (50-80V; up to 300V for plasma cutters, though that may be only during ignition, I forget), and, at available current and ~1atm ambient pressure, this limits arc length to less than a cm.
Or why you see thousands of volts dropped across long glow or arc discharges, even rather hot ones like the above video.
Which, I have no idea why protection didn't fire in that case, maybe it wasn't equipped, maybe it wasn't operational, maybe the lines are designed for as much current (and thus the switchgear at either end as well) and there was enough line resistance (and reactance!) to limit current below the fusing threshold. Would have to look up the particular case and see if there was any analysis or a report on it. Such incidents are fairly rare, highly visible (for obvious reasons, lol) and tend to make the news, at least locally, so there's probably something out there. I'm not interested enough to look it up myself but anyone who wants to give it a try is welcome to do it -- research is good exercise.
In any case, this is why I've been trying to emphasize the collective effect of myriad minute discharges. A corona discharge isn't a static effect, it's highly dynamic, far more dynamic than you can see by eye, probably than you can see by any high-speed camera even. It looks like a quasi-static or chaotic discharge by eye, but this is simply because your eyes can't resolve billions or trillions of discharges per second.
The power dissipated by those myriad discharge paths, must necessarily come from the energy spent charging and discharging the air surrounding the breakout. There's no other energy flow, and current is limited by the reactance of that surrounding air.
And neither can your circuit [resolve these myriad paths]; nor does it care. All it cares about is load current at the fundamental, and harmonics if relevant. Each individual discharge lets off a microscopic EMP burst, contributing to radiation from MW to microwaves -- which is how corona and arc discharges are so noisy, they emit a ton of RF noise. But averaged over a cycle of say 40kHz, it doesn't amount to much, and just looks like a resistor load with excess noise.
Air breaks down because the maximum field strength is exceeded, electrons are torn from molecules, and avalanche discharge ensues. It doesn't matter where this happens, it can happen at random in free air, but it's most likely to occur near an electrode (at a point or corner, a sharp curvature, maximum field strength at the surface), or extending from a discharge which now pierces into the surrounding low-voltage field, making its own needle tip and so on. Streamers shoot and branch as they fill into nearby space, charging it up, making it unipotential to the electrode, at least momentarily until the voltage reverses and the whole process repeats anew.
But none of this is material to the electronics, all that's needed is a representative RC load at the driven fundamental frequency. Most likely, the transformer will have considerable leakage inductance and secondary stray capacitance, making its resonance quite strong, and a full-wave inverter driving that is quite effective, the leakage acting to filter harmonics -- which validates our earlier assumption of single-frequency analysis, and which without, we might want to choose a different analysis (a more detailed RCRC load equivalent, perhaps, and a square-wave source).
Tim