CODEWITHSUNDEEP
pythonoptimizationbooleancvxpy
Maxime LIBSIG
Maxime LIBSIG

Reputation: 21

Constraint on a boolean variable?

I am trying to implement a unit commitment problem with cvxpy with a rampup cost. However I am struggling to identify the ramp-ups. (when variable switch from 0 to any positive value) In other terms, I have a variable g, positive and I want to determine when g is 0, or when g>0. I know it's often easier with some tolerance, that's not really an issue for me.

Below is some code to put this in context :

import cvxpy as cp
n=50    

g=cp.Variable(n)
ispos=cp.Variable(n,boolean=True)

I would need a constraint that looks like this :

I tried to work around something like below, but I cannot get a syntax CVXPY is happy with.

constraints=[ispos == (g>=0)]

I get the following error

Exception: Cannot evaluate the truth value of a constraint or chain constraints, e.g., 1 >= x >= 0.

Also tried with this :

constraints+=[ off[t]+(1 if g[t]<=0 else 0)== 1 for t in range (n)]

But I get the following error:

Exception: Cannot evaluate the truth value of a constraint or chain constraints, e.g., 1 >= x >= 0.

Upvotes: 1

Views: 1640

Answers (1)

Erwin Kalvelagen
Erwin Kalvelagen

Reputation: 16782

You want:

ipsos[t] = 0 ==> g[t] >= 0
ipsos[t] = 1 ==> g[t] <= -0.001

This can be modeled as:

g[t] >= L[t]*ipsos[t]
g[t] <= -0.001*ipsos[t]+U[t]*(1-ipsos[t])
g[t] ∈ [L[t],U[t]]

I assume here that the lowerbound L[t] is a negative number. In practice, I would drop this -0.001 thing and just do:

g[t] >= L[t]*ipsos[t]
g[t] <= U[t]*(1-ipsos[t])
g[t] ∈ [L[t],U[t]]

Upvotes: 0

Related Questions