PI Tunning & Reference Tracking With Genetic Algorithm

Hi Friend,

I want to optimising PI parameters(Kp&Ki) with Genetic Algorithm.

in simulation that you can see http://projeharshad.persiangig.com/image/Untitled-1.jpg range of the Kp&Ki that calculate from equations is very different. for example Kp&Ki in PI(1) is 3031059 & 415464655 and for Kp&Ki in PI(2) is 7.3 & 2.66. Can I use the Genetic Algorithm for PI tunning for Reference Tracking in this system that range of the Kp&Ki in PI's very different?

Answer 1

[Alan Weiss]

You can set the initial range of the genetic algorithm to anything you like, as explained here. This is an important step whenever you use the genetic algorithm.

You will probably get better results if you scale your problem a bit, such as taking the log of the PI(1) term, the one you said is perhaps 3031059 to 415464655. Well, perhaps I misunderstand, but I'm sure you can do some scaling to get the parameters to be in a reasonable range.

But allow me to make my usual suggestion that you do not use the genetic algorithm, and instead try patternsearch. patternsearch is faster, more robust, and easier to tune. The only difference in syntax is that you must supply a starting point for patternsearch. This is probably very easily done, especially if you scale the problem.

Good luck,

Alan Weiss

MATLAB mathematical toolbox documentation

[MHz]

Thanks Alan,but I told that parameters range of 2 PI very different,in other word PI paramaeters in PI(1) is 3031059 & 415464655 and PI paramaeters in PI(1) is 7.3 & 2.66.also number of my variables are 4. in this condition what is the start point of pattern search?

[lithika]

can u tel me how to get the error in time domain (ISE) so that PI tuning optimisation using GA can be done...