Genetic Algorithms (Error is too big)

Greetings , i'm doing a parametric optimization for the induction machine ,i've made a simulink model and a mathematical model for the implentation on the GA toolbox, the function works fine when i excute the optimization from the toolbox , but the results are just awfull (never gone under 300) , knowing that i'm minimazing the function , a such results is way out of the acceptable , any thoughts on how to get better results ? , cuz i've been fiddling with it since a relatively long time and it doesn't seem to work

if you need to see the programmes , just comment here and leave your email or something and i'll send them to you , plz i really need help with this for my project

[Walter Roberson]

http://www.mathworks.com/matlabcentral/answers/7924-where-can-i-upload-images-and-files-for-use-on-matlab-answers

[Lotfi]

thank you Walter, well here is the link http://www.crocko.com/2F6FADD44D9049E18605B2C2EFAAF4F8/Mas.rar 3 files are contained , file.m will get the results of simulation of MAS , and machgene contains the function where i solved the induction machine equation with runge kutta 4th order (solved equation of the model = calculated values), which means the one which is gonna be called by ga toolbox for the criteria minimization

[Walter Roberson]

Note: several of the volunteers cannot handle .rar files. (On the other hand the ones I know of also do not have Simulink or the GA toolbox)

[Lotfi]

Ga toolbox is integrated in all new or relatively new matlab versions ,same thing for simulink

[Walter Roberson]

No they are not.

>> version

ans =

7.14.0.739 (R2012a)

>> simulink
Undefined function or variable 'simulink'.

>> ga
Undefined function or variable 'ga'.

Are you perhaps thinking of the Student Version?

[Lotfi]

no i'm not using the student version, i'm using the standard 2009 version >> version

ans =

7.8.0.347 (R2009a)

i've used the 2012a version on another pc once, and as far as i recall , the toolboxes are a bit messed up but the simulink works fine , anyway , can you help me on this matter , i mean how can i demonstrate my problem here if nobody has simulink (for the ga , you can type help ga and you'll get a satisfactory explanation on how to use it with a toolbox : X = GA(FITNESSFCN,NVARS) ....etc) ? thank you in advance Walter

[Walter Roberson]

I have R2012a, the full-priced commercial version purchased only a few weeks ago, and it does NOT include Simulink or the Optimization Toolbox, or any of the other MATLAB toolboxes. The particular PC you used might have happened to have those installed.

Some of the volunteers do have Simulink and also the Global Optimization toolbox. I would have difficulty in remembering exactly whom, though. I do not recall any of the 50 most active volunteers as happening to having experience with both, but I don't read everything here (just most things.)

Historically, most questions about GA go unanswered. That's just the way it goes with volunteers: sometimes people just do not have the resources or experience or the interest to answer.

Answer 1

[Alan Weiss]

If you have GA, then you also have PATTERNSEARCH. I recommend PATTERNSEARCH as a more reliable, faster algorithm.

The only difficulty using PATTERNSEARCH is generating initial points. If you have bounds for all variables, you can use

 x0 = lb + rand(size(lb)).*(ub - lb);

You can also experiment with tuning PATTERNSEARCH more easily than tuning GA. For example, try different poll methods.

Alan Weiss

MATLAB mathematical toolbox documentation

[Lotfi]

Thank you Alan , but i'm gonna have to do this one with genetic algorithms , it's for my graduation project