1

I need your help in the following optimisation problem. I have a maximisation Mixed Integer linear programming problem. I would like to consider the minimum & maximum fixed fee.

I've done it in this way.. 

cost = max(minimum fixed cost , cost rate * x)
cost >= minimum fixed cost
cost >= cost rate * x
cost = min(maximum fixed cost , cost rate * x)
cost <= maximum fixed cost
cost <= cost rate * x

But, This turns infeasible solution. Would you please help me in optimising such a problem.

Mohammad
  • 21,175
  • 15
  • 55
  • 84
user7090205
  • 21
  • 2

2 Answers2

2

piecewise linear functions

I think what you mean is the following:

Let

A = minimum fixed cost / cost rate
B = maximum fixed cost / cost rate

Then you want to model the piecewise linear function:

cost = minimum fixed cost   if x < A
       cost rate * x        if A <= x <= B 
       maximum fixed cost   if x > B

enter image description here

Using piecewise linear functions inside a MIP model is not a problem. You can do this by different approaches:

  • using extra binary variables (see (1))
  • using SOS2 variables (see (1))
  • systems like AMPL and Gurobi have special facilities to express piecewise linear functions.

Example formulation

A formulation with SOS2 variables can look as follows:

Introduce data points px and py

  px     py
  --------------------------
  0      minimum fixed cost
  A      minimum fixed cost
  B      maximum fixed cost
  C      maximum fixed cost

where we assume 0<=x<=C. I.e. C is an upper bound on x.

Then do:

  set p = {1,2,3,4}
  sos2 variables lambda(p)
  sum(p, lambda(p)) = 1
  x = sum(p, lambda(p)*px(p))
  cost = sum(p, lambda(p)*py(p))

See e.g. (2)

What is wrong with your approach

Note that your approach (shown in the question) is incorrect:

cost >= minimum fixed cost
cost >= cost rate * x
cost <= maximum fixed cost
cost <= cost rate * x

is really

minimum fixed cost <= cost <= maximum fixed cost
cost = cost rate * x

which limits x to A <= x <= B.

References

(1) H.Paul Williams, "Model Building in Mathematical Programming", Wiley

(2) GAMS: Piecewise linear functions with SOS2 variables

Erwin Kalvelagen
  • 15,677
  • 2
  • 14
  • 39
  • Thank you very much for your answer. However, I don't want to limit x. x can be greater than B or smaller than A. If x is greater than B, maximum fixed cost should apply. If x is less than A, minimum fixed cost should apply. I went through the problems presented in chapter 9 in the book but still I can relate them to my problem. Would you please provide me with an example of how such a problem is solved.I would greatly appreciate it. – user7090205 Apr 06 '17 at 15:23
  • You formulation imposes these limits on x. It is disappointing you don't understand the Williams book. It is considered a very good and quite accessible. I have added a SOS2 example. – Erwin Kalvelagen Apr 06 '17 at 22:04
  • Thank you again @Erwin Kalvelagen. I've grasped the concept of SOS2. However, I'm still getting an unfeasible solution in my model :( Actually, my variable X looks like this x(a, b, c, d, e). So for each x(a, b, c, d, e) I've created variables x1(a, b, c, d, e), x2(a, b, c, d, e), x3(a, b, c, d, e), x4(a, b, c, d, e). Similarly, for the binary variables for each line segment y1(a, b, c, d, e), y2(a, b, c, d, e), y3(a, b, c, d, e) to control adjacent x. – user7090205 Apr 07 '17 at 18:22
1

Piecewise linear function:

cost = 0.02 x  if 0   <= x <= 500
       0.03 x  if 500 <= x <= 1500 
       0.04 x  if 1500<= x <= 10000

SOS2 Solution:

x[a, b, c, d, e] = (x1[a, b, c, d, e] * 0)    + 
                   (x2[a, b, c, d, e] * 500)  + 
                   (x3[a, b, c, d, e] * 1500) +
                   (x4[a, b, c, d, e] * 10000) 

cost[a, b, c, d, e] = (x1[a, b, c, d, e] * 0       * 0   )+ 
                      (x2[a, b, c, d, e] * 500     * 0.02)+
                      (x3[a, b, c, d, e] * 1500    * 0.03)+
                      (x4[a, b, c, d, e] * 10000   * 0.04)

x1[a, b, c, d, e] + x2[a, b, c, d, e] + x3[a, b, c, d, e] + x4[a, b, c, d, e]>= 0

x1[a, b, c, d, e] <= y1[a, b, c, d, e]
x2[a, b, c, d, e] <= y1[a, b, c, d, e] + y2[a, b, c, d, e]
x3[a, b, c, d, e] <= y2[a, b, c, d, e] + y3[a, b, c, d, e]
x4[a, b, c, d, e] <= y3[a, b, c, d, e]    

y1[a, b, c, d, e] + y2[a, b, c, d, e] + y3[a, b, c, d, e]  = 1

I've done it in this way. However, the solver still shows unfeasible solution. Can you see something wrong in this formulation?!

user7090205
  • 21
  • 2