lower bounds of variables given by iterating over dictionary instead of matrix

Answered

Hussein Sharadga

March 14, 2023 02:08 Edited

• 
• 
• 

I have a large problem with binary variables.

The binary variables sometime should be one. This is given by the lower bound on this binary variable.

The normal condition: lower bound=0, upper bound=1

Must-equal-1: lower bound=1, upper bound=1

 

Gurobi violates the lower bound for the Must-equal-1 case: It returns a value of zero sometime. That happens if I store the lower bound and upper bound as a dictionary. For example, I have 100 (n_j=100) devices and 5 time steps (n_t=5).

The dictionary keys will be numbers between 0-99. For example, the lower bound dictionary for device 0 will be like this:

B_lb[0]=[0,0,1,0,0]

 

This is how I add the binary variable with the bounds to my model:

B = m.addVars(n_j,n_t,vtype=GRB.BINARY,lb=B_lb, ub=B_ub) 

 

 If I convert my dictionary to a matrix, it works fine?

 

 

 

 

 

Related to

• Python

0

• 

  Riley Clement

  • Gurobi Staff

  March 14, 2023 03:02

  Hi Hussein,

  Your B_lb dictionary has the following structure:
  keys = single indices for device
  values = list of lower bounds at each timestep

  But it needs to be structured like so (as per the addVars docs):
  keys = 2-tuple of (device,time step) pairs
  values = a single number, the lower bound, for the corresponding device and time step

  You can convert the former into the latter with the following code

  B_lb_new = {
      (device,time_step):lb 
      for device,lb_per_timestep in B_lb.items()
      for time_step,lb in enumerate(lb_per_timestep)
  }

  - Riley

  1

Please sign in to leave a comment.