I have a mixed integer non linear model and now I am trying to reduce the number of nonlinear constraints because the model is very big and I keep getting infeasibilities or errors.
Two of the constraints that are non linear are as following:
const11(n).. p(n) =e= 1 - (1 + (d(n) / 229)) ** (-0.2);
const12(n).. il(n) =e= p(n) * y(n) ;
where d(n) and y(n) are choice variables.
I tried using grid point approximation but since the constranits are not seperable I had a hard time in grid point approximation.
Does anyone have any idea how to change these two nonlinear constraints into linear constraints?
I appreciate your help.
Problems with modeling
3 posts • Page 1 of 1
Without further knowledge about the type of variables, the size of n etc it's hard to recommend anything. The constraints are nonlinear and while there are ways to linearize something like this (e.g. piecewise linear approximation) it's hard to suggest something useful without more details.
For linearizing the nonlinear equations, you can use piecewise modulation. Others, like you are getting an infeasible solution, try to check the boundaries of your variables. Your solution might be out of the feasible region.