GA Optimisation, the final output solution violate the constraints.

I am using GA for minimising power loss in a electrical distribution network subjected to inequality constraints. I need to determine 120 decision variables and all of them should be integers. After the GA stops optimisation, the final Fval and penalty values are different and the constraints are also violated. Why is it so and How am i supposed to overcome this problem?

Answer 1

[Alan Weiss]

You are probably exceeding the limits of what ga can do. You have a large constrained nonlinear problem with integer constraints, and these problems are hard. If ga cannot find a feasible point, then you can try giving it a much larger population size (say 1e4 or larger), but then it will run very slowly, and you are still not guaranteed of getting a feasible answer.

Good luck,

Alan Weiss

MATLAB mathematical toolbox documentation

[roshani thapa]

Thanks Alan, i set the population size to 10000 and it worked. I had to wait for 4 days to obtain the output with intel core i5 processor and 8gb RAM. But i found the penalty value plot was broken after few generations and continued from a different point. what does this mean?

[Alan Weiss]

The generations up to about 130 had no feasible points, Then a feasible point was found, and the optimization continued.

Alan Weiss

MATLAB mathematical toolbox documentation