Description

This model takes some scenario data (that comes from a true tree),
builds a fan and lets ScenRed construct a tree. One without loss of
information, the other with some reduction requirement.

Small Model of Type : GAMS

Category : GAMS Model library

Main file : srtree.gms

$title Simple Scenario Tree Construction Example (SRTREE,SEQ=390)

$onText
This model takes some scenario data (that comes from a true tree),
builds a fan and lets ScenRed construct a tree. One without loss of
information, the other with some reduction requirement.


H. Heitsch, W. Roemisch, and C. Strugarek
Stability of Multistage Stochastic Programs
SIAM Journal on Optimization 17 (2006), 511-525

Keywords: stochastic programming, scenario tree construction, GAMS ScenRed
$offText

Set
   s 'scenarios'   / s1*s4 /
   t 'stage'       / t1*t5 /
   r 'random data' / r1*r4 /;

Parameter p(s) 'probability' / (s1,s2) 0.25, s3 0.3, s4 0.2 /;

Table sdata(s,t,r)
            r1    r2    r3     r4
   s1.t1  42.5   9.1   7.5  120.0
   s1.t2  39.8  11.2   8.4   90.0
   s1.t3  37.6  14.0   6.3  110.0
   s1.t4  38.9  12.4   8.1  130.0
   s1.t5  40.3  14.9   7.2  120.0
   s2.t1  42.5   9.1   7.5  120.0
   s2.t2  39.8  11.2   8.4   90.0
   s2.t3  37.6  14.0   6.3  110.0
   s2.t4  38.9  12.4   8.1  130.0
   s2.t5  38.4  15.2   8.9  100.0
   s3.t1  42.5   9.1   7.5  120.0
   s3.t2  39.8  11.2   8.4   90.0
   s3.t3  37.6  14.0   6.3  110.0
   s3.t4  35.7  13.8   7.5  120.0
   s3.t5  37.6  14.9   9.3   80.0
   s4.t1  42.5   9.1   7.5  120.0
   s4.t2  39.8  11.2   8.4   90.0
   s4.t3  37.6  14.0   6.3  110.0
   s4.t4  35.7  13.8   7.5  120.0
   s4.t5  36.3  12.8  10.3   90.0;

* Construct a fan
$eval nnodes card(s)*(card(t)-1)
Set
   n        'nodes' / n0*n%nnodes% /
   nn(n)    'next nodes'
   anc(n,n) 'ancestor relation';

Parameter
   prob(n) 'node probability'
   rv(r,n) 'random values';

nn('n0') = yes;
loop(s,
   loop(t$(ord(t) < card(t)),
      loop(nn(n),
         if(sameas('t1',t),
            anc(n+1,'n0') = yes;
         else
            anc(n+1,n) = yes;
         );
         prob(n+1) = p(s);
         rv(r,n+1) = sdata(s,t+1,r);
      );
      nn(n) = nn(n-1);
   );
);
prob('n0') = 1;
rv(r,'n0') = sdata('s1','t1',r);

* Initialize ScenRed
$set srprefix test
$libInclude scenred

Set anc_noloss(n,n), anc_red(n,n);

Parameter prob_noloss(n), prob_red(n);

* Scenred call
ScenredParms('red_percentage') = 0;
$libInclude scenred %srprefix% tree_con n anc prob anc_noloss prob_noloss rv
abort$(card(anc_noloss) <> 8) 'scenred tree construction gave incorrect noloss tree', anc_noloss;
display anc_noloss;

* Scenred call
ScenredParms('red_num_leaves') = 3;
$libInclude scenred %srprefix% tree_con n anc prob anc_red prob_red rv
display anc_red;