I think the first step in solving the problem is to determine the response rates of the various signals. Firstly the current sensor - what sort of response delay does this have? As for the thyristor control - we can only change this on a per half cycle basis (preferably on a full cycle as we don't want to introduce too much assymetry into the transformer). Is the thyristor timing fine enough to give you the control you require? If you have only control in 10% steps the best output you could expect is +/-10%!

If your current sensor is quite fast (less than a cycle in response time) you may be able to use a simple proportional control rather than P.I.D. What I'm trying to say is not to get too hung up about PID until you've tried some simpler control methods. In trying something simpler you may either solve your problem or gain a better insight into the problem so that you can try more sophisticated control solutions.

I needed to design a positioning system for a large display board. The first thought was to have an optical encoder and a sophisticated servo motor drive - then I started looking at the costs! I only wanted fairly crude positioning (+/-2degrees) so I ended up using a pot for position feedback into a 8bit a/d and some simple code to do a window comparision and output either forward or reverse to a cheap variable speed drive. The VSD had accell/decel ramp so that provided a bit of integral term and when my software detected the required position, it would stop. So what I ended up with was an ID loop (integral and difference) realised in a few lines of 'c' code. So the moral here is, start simple then move to more complex if required.