I see, but is there any advantage using a transformer instead of a opto-coupler, like they've done in the attached circuit. Is this a possible solution?
No, that circuit would be sketchy at 20A, much less 1kA, mainly because the energy delivered to the gate will depend on when in the cycle triggering occurs. This is fine for a static switch, and fine for phase control at low current (ie - under 20A), but not okay for large thyristors that might need as much as 1A at 3-9V for <10us to reliably trigger, but which can't tolerate that same voltage/current for the entire cycle without exceeding the gate power rating.
If you've never worked with high power thyristors before, well... I'm not one to tell people what they can't do - who am I to judge? - but let's just say the learning curve is steep with these devices (that said, they are easier to use than GTOs!)
Regarding dI/dt. Say the thyristors are rated at dI/dt = 300A/µs. What value should this be compared to?
This is pretty straightforward: use the inductor equation to figure out how much inductance is needed to limit dI/dt given the peak value of the mains voltage. Since a saltwater load is effectively a low-inductance resistance, you will almost certainly need additional inductance here; probably somewhere in the range of 2-4uH, depending on how fast the SCR turns on (generally 1-4us for phase control SCRs; <1us for inverter SCRs). You will need an RC snubber, with component values more likely in the range of 1uF and 1-10 Ohms, rather than the suspiciously generic-looking 0.1uF and 100 Ohms shown in the above schematic.
Another note - electrolysis of saltwater is how bleach (NaOCl) is made industrially, though with DC, of course. Now, with AC the cathode and anode switch places 50 or 60 times per second, so ion migration can't really happen, but I do know that gases are produced in water resistors even when AC is applied, so beware. You might find it is altogether preferable to submerge coils of stainless steel wire rope into a barrel of deionized water to use as a water-cooled resistor, rather than use salt water itself as the resistor. Also, the coils of wire rope will definitely have enough minimum inductance.