I'm using PSO to tune the parameters of the PID controller . my model uses a s-function block to represent the equations of the system in SIMULINK. When I try to tune the system using PSO offline, the system takes very long time (almost half an hour) with no response, and I stopped the run. After all this time the program was still calculating the s-function for flag = 1 at time 9.1 and PSO didn't even run!!
How can I solve this delay problem ?
Dear Eman El-Gendy,
Have you tried setting the PID parameters using conventional methods such as Ziegler and Nichols or even manually.
Regarding the PSO part, usually it does not take much time, in my opinion it will be necessary to see if there is no problem with the PSO algorithm itself.
Best regards
Have you checked he accuracy of your matlab codes? Maybe you have done a mistake there that has created a loop or .... that makes infinity cycles.
however, what is your computers processor ?
I used the conventional ZN method for PID tuning and the system worked well .. then I used fuzzy gain scheduling and worked well too .. the problem is with the PSO code and I don't know why ..
The processor is i7 with 8 gb ram
Since each trial of the PSO requires simulation of the model, problem may be caused by inapropriate selection of the Solver and/or maximum step size.
If Your model is relatively simple (SISO) you should consider the repleacement of it by transfer function or state space model and simulation without simulink (by using lsim command).
Long simulation time may also be casued by unstable response (with chattering in signals).
Regards,
Tomasz
PSO doesn't take long time. The convergence speed of PSO is faster as compared to the other conventional AI technique i.e. GA, NRM etc. The proper selection of constriction factor and weighing factor is very important. The proper tuning of Kp, Ki and Kd also depends upon the cost function.
I think that, the PSO Code is the main problem and it doesn't match your goal clearly. There must be a thin difficult point which shoots the convergence environment out of the space. There is an unreasonable part in the code. Regarding my experience, try to check the code and match it again with the mathematical concept which you follow.
Don't worry, this is a conventional trouble with coding.
check your stopping criterion also. you may set it as when maximum number of iterations are reached.
It heavily depends on the problem you're trying to optimize (mathematical model), the way you wrote your program and the specs of your computer. Try to keep your model as simple as possible.
how can add my pso program to simulink block for tunning of Kp and Ki for shunt active filter
- Badges
- Science topic