How can I set nonlinear constraints in a nonlinear grey-box model?

Hi,

I´m estimating pramaters of a nonlinear state-space model with the System Identification Toolbox and the function nlgreyest.

Is it possible to add nonlinear constraints to the estimation?

For example:

One equation of my state-space model is

dx(1) = K1*x(1) + K2*x(2) + K3*u(1);

And now I want to add a nonlinear constraint like this to the estimation.

0.5 = K1 / (2*sqrt(K2));

Can someone help me solve this problem?

Thanks, Philipp

[soeren graabaek]

Hey,

Did you find a solutuion?

Best regard

Søren

Answer 1

[Rajiv Singh]

Linear or nonlinear constraints are not supported in nlgreyest directly. Depending upon the problem, you can sometimes get away with a re-parameterization, and.or simple min/max bounds. For example in the example you have given, you can eliminate K1 by replacing it with sqrt(K2), and a lower bound K2>0.