Optimization using fmincon for a simulink model

Hi, I have a simple simulink model using embedded matlab function. The model is supplied with a signal (T) which is a 100 by 1 column array, and another input variable (N) which is a constant value. The model produces two outputs (P &D) which are of the same size as T. The equation for the model is D=T/(P*N)
By providing the values of different N (one value for one simulation), D can be changed. Now I want to minimize D, for a range of values of N between 1 and 250. I have specified my objective function as function [D]=myobjectivefcn(x) D=T./(P.*x) end
I am running the optimtool to select the solver and specify the upper and lower bound for x as 1 and 250, with sart value as 1. But its not working. Can somebody please help me, I dont know how to connect my simulink model with the optimization. Really really appreciate any help in correcting the code.
Thanks

2 Kommentare

Khalid
Khalid am 25 Nov. 2014
I have a very similar question, as does another poster from this thread:
Is moving over to Matlab really the best option? Matlab is great in terms of how flexible it is; but I find I am better at re-using parts of code/model and finding my mistakes when using Simulink.
Besides, this must be possible, as optimisation from within a Simulink model is completed in the built-in MPC controller block (as well as in other places I'm sure). Out of curiosity I had a quick look at how they do this, and there is a MATLAB Function block, and within that block they make a call to "mpcblock_optimizer_mex", which is set as extrinsic using
coder.extrinsic('mpcblock_optimizer_mex');
But unfortunately as it's a mex function I can't see any more than that. Any thoughts anyone? As I've noted in the other post I tried declaring fmincon as an extrinsic function but its then not possible to pass function handles to it.
Any pointers much appreciated.
- Khalid.
Yousif Al-sagheer
Yousif Al-sagheer am 23 Apr. 2015
Bearbeitet: Yousif Al-sagheer am 23 Apr. 2015
Hi, MPC block in simulink is designed for linear system. In linear system predictive control the control low finds the optimal values using a control law equation which consists of some matrix multiplication (the control law can be found in literature, "Model based predictive control: A practical approach J. A. Rossiter P.62"). So I do not think MPC block in simulink contains any fmincon command inside it.

Melden Sie sich an, um zu kommentieren.

Antworten (1)

Babak
Babak am 29 Nov. 2012

0 Stimmen

My suggestion, move everything in MATLAB script language (m-files) if possible and don't do an optimization in Simulink. You can write your own time steps like simulink in script.
If you cannot move everything to script lamguage: I would write an m-file and put a for loop, looping on different values for N (or x) and inside the for loop call your simulation to run, with sim() command. you can get the result out of your simulink model in different ways, like you could use a "to workspace" or "to file" block. Before, the end of the for loop, you need to save the output result of simulink, somehow, like in a local matrix in the main calling script file.

Kategorien

Mehr zu Manual Performance Optimization finden Sie in Hilfe-Center und File Exchange

Gefragt:

am 29 Nov. 2012

Bearbeitet:

am 23 Apr. 2015

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by