Dear Aissam Riad Meddour,

I suggest you to see link and attached file i hope that references are on topic.

Conference Paper Using genetic algorithms in design optimization of the flux ...

Best regards

Dea Aissam,

Indeed, heuristic optimization algorithms and genetic algorithm is no exception are great tools to run an optimization and Their main advantage is that you can get a global solution instead of stuck on a local one. Unfortunately, all of this comes with an extreme time-consuming especially in the case of big models such yours. To go around this problem I would suggest you use reduce order techniques to make your model smaller and as a result, reducing the computational time.

Cheers,

Kouider

Dear Kouider Bendine

Thanks for your answer,do you suggest to start with a big number of population and hope that it coverge in two or three generations or use a little number of pop hoping it converge a lot gaster?

Thanks

Aissam Riad Meddour the problem is not just the population number, there is also the generation number the mutation, and crossover rate so as a start I would recommend you to use a law population number in order to check the behavior of the algorithm and based on u can increase the number. also it will be also great if u set a convergence criteria to control the generation number used.

Dear Aissam Riad Meddour,

Genetic algorithm is recommended for discrete time systems. But for continuous time systems, the PSO algorithm is better.

Therefore, my suggestion is to choose one of the two according to the system under study. Also to develop it, the performance of both algorithms can be compared with each other.

Compare GA results with MATLAB Pattern Search (PS). At the optimization of the dynamical system, in our case delay chain, the Pattern Search provided better performance. Difference between Pattern Search and GA approach in MATLAB are briefly explained in the paper ( not on the case of Motor - FEA but applying GA & PS on Dynamical Model defined in state space ): Article Determination of recruitment and transition strategies

In MATLAB, usage of GA and PS is similar, so with small effort two approaches could be compared.

It would also be good to use MATLAB parallel computing functionality.

Besides, it would be worth checking Differential Evolution approach: Article Selective Pressure Strategy in differential evolution: Explo...

Even you can prefer a metaheuristic algorithm that is not population-based to avoid the time computation; Example: Simulated Annealing and its variants.