Originally posted by: StefanWeiser

Hey,

I have a model that includes a price px that depends on a chosen amount x [Periods][Year] which is my decision variable.

I succesfully implemented the piecewise function for x[Period] and only one year. 

dexpr float CostPerPeriode [t in Periods] = 
piecewise ( i in 1..N){
     s[i] -> p[i];
     s[N+1] 
 } (0, 0) x[t];

My goal is to extend the model to several years, all with the same periods/month. The model should remain the same and the price px should also just depend on the amount choosen per periode. 

With my extended model a error message says that  the piecewise function is not an array type and does not work with two parameter variables. 

Do you have an other idea how to solve my problem? 

regards

Hi,

let me give you a small example:

range Periods=1..12;
int N=5;

dvar int x[1..4][Periods] in -100..100;

int s[y in 1..4][i in 1..N]=-i;
int p[i in 1..N]=i;

dexpr float CostPerPeriode [y in 1..4][t in Periods] =
piecewise ( i in 1..N-1){
     s[y][i] -> p[i];
     s[y][N]
 } (0, 0) x[y][t];
 
 subject to
 {
 CostPerPeriode[1][2]<=10;
 }

regards

Originally posted by: bp325

Agreed with Alex on the solution, but I created a very similar problem (in which the piecewise curve shapes vary for each time period) and get a strange result that I can't explain...

I have:

dexpr float ElecCost[p in Period]
  Eprice[p]*(piecewise(l in Line){
   M[p][l] -> X[p][l]; 0
   }(0,0)Gen[p]);
 
 dexpr float TotCost = sum(p in Period) ElecCost[p];

which is fine, except that the values of ElecCost are different before being used in the summation and within the summation. The sum of the latter is ~80% greater than that of the former when they should be exactly the same!

Any thoughts on why this might be?

Also, sorry for hijacking the post, but the subject fits (this weirdness isn't happening with other single variable piecewise curves), so thought it would be good to have everything in one place.

Best,

Bryn

Hi,

can you attach a .mod so that other users can try ?

regards

Originally posted by: bp325

OK, I've spent time slimming it down to a simple MOD file that shows the problem, see attached. You'll notice that TotCost1 is not the sum of ElecCost1, even though it is set to be!

Best,

Bryn

Hi,

indeed this looks strange.

Can you try with dvar instead of dexpr ?

dvar float ElecCost_1b[p in Period];
 dvar float TotCost1b;

and then

minimize TotCost1b + TotCost2 + EnPurchased;
 
 //Constraints
 subject to {
 
 forall( p in Period) ElecCost_1b[p] ==
  E_price[p]*(piecewise(l in Line){
   M_e1[p][l] -> X_e1[p][l]; 0
   }(0,0)Gen[p][1]);
 
 
 TotCost1b ==  sum(p in Period) E_price[p]*(piecewise(l in Line){
   M_e1[p][l] -> X_e1[p][l]; 0
   }(0,0)Gen[p][1]);

regards

Originally posted by: bp325

Thank you Alex, that seems to work. So, if creating a piecewise function with two parameters, it is best to have the function as a decision variable rather than decision expression. Would be good to work out why, but for now I'm happy that a solution is possible :)

Bryn