GLPK Hydro Storage - Infeasibility

Question

I am trying to fiddle around with GLPK and the mathprog language. I am trying to implement a simple model for a hydro storage (just turbining, no pumps). But I am getting infeasibilities.

Pouring over the initial conditions, this should be feasible.

---

Here's the code:

set T;

#========================================================
# Time Series

# Price
param price{i in T};
# Inflow
param inflow{i in T};

#========================================================
# Unit description

param release_max>=0;
param release_min>=0;
param fill_max>=0;
param fill_min>=0;
param fill_start>=0;
param fill_end>=0;


#========================================================
# optimization variables
var release{i in T}>=0;
var fill{i in T}, >=fill_min, <=fill_max;

# objective: Maximize profit
maximize obj: sum{i in T} price[i] * release[i];
s.t. fill_current {i in T: i>1}: 
        fill[i] = fill[i-1] - release[i] + inflow[i];
s.t. fill_con_start {i in T: i=1}:
        fill[i] = fill_start;
s.t. fill_con_end {i in T: i=card(T)}:
        fill[i]>=fill_end;

solve;
data;

param release_max:=100;
param release_min:=0;
param fill_max:=10000;
param fill_min:=0;
param fill_start:=5000;
param fill_end:=5000;

param : T :   price    inflow :=
         0     15     150
         1     17     170
         2     18     180
         3     22     220
         4     55     550
         5     40     400
         6     65     650
         7     10     100
         8     12     120
         9      4     40
;
end;

---

Here's the output:

GLPSOL: GLPK LP/MIP Solver, v4.55
Parameter(s) specified in the command line:
 --cover --clique --gomory --mir -m Hydro_test.mod
Reading model section from Hydro_test.mod...
Reading data section from Hydro_test.mod...
58 lines were read
Generating obj...
Generating fill_current...
Generating fill_con_start...
Generating fill_con_end...
Model has been successfully generated
GLPK Simplex Optimizer, v4.55
10 rows, 19 columns, 35 non-zeros
Preprocessing...
PROBLEM HAS NO DUAL FEASIBLE SOLUTION
Time used:   0.0 secs
Memory used: 0.1 Mb (102683 bytes)
>Exit code: 0    Time: 0.316

Can anybody help me out?

Answer 1

Stupid is what stupid does:

I forgot to limit the release parameter and there were some problems with the index (e.g. initial fill level assigned too late etc.). Therefore the problem was unbounded

The objective should look like this (see last line):

# objective: Maximize profit
maximize obj: sum{i in T} price[i] * release[i];
s.t. fill_current {i in T: i>0}: 
        fill[i] = fill[i-1] - release[i] + inflow[i];
s.t. fill_con_start {i in T: i=0}:
        fill[i] = fill_start;
s.t. fill_con_end {i in T: i=card(T)-1}:
        fill[i]>=fill_end;
s.t. release_current {i in T}:
        release_max>=release[i]>=release_min;