clearlaksp.gms : Scenario Reduction: ClearLake exercise

**Description**

Exercise, p. 44: The Clear Lake Dam controls the water level in Clear Lake, a well-known resort in Dreamland. The Dam Commission is trying to decide how much water to release in each of the next four months. The Lake is currently 150 mm below flood stage. The dam is capable of lowering the water level 200 mm each month, but additional precipitation and evaporation affect the dam. The weather near Clear Lake is highly variable. The Dam Commission has divided the months into two two-month blocks of similar weather. The months within each block have the same probabilities for weather, which are assumed independent of one another. In each month of the first block, they assign a probability of 1/2 to having a natural 100-mm increase in water levels and probabilities of 1/4 to having a 50-mm decrease or a 250-mm increase in water levels. All these figures correspond to natural changes in water level without dam releases. In each month of the second block, they assign a probability of 1/2 to having a natural 150-mm increase in water levels and probabilities of 1/4 to having a 50-mm increase or a 350-mm increase in water levels. If a flood occurs, then damage is assessed at $10,000 per mm above flood level. A water level too low leads to costly importation of water. These costs are $5000 per mm less than 250 mm below flood stage. The commission first considers an overall goal of minimizing expected costs. This model only considers this first objective.

**Small Model of Type : ** SP

**Category :** GAMS EMP library

**Main file :** clearlaksp.gms

```
$TITLE Scenario Reduction: ClearLake exercise (CLEARLAKSP,SEQ=72)
$ontext
Exercise, p. 44:
The Clear Lake Dam controls the water level in Clear Lake, a
well-known resort in Dreamland. The Dam Commission is trying to
decide how much water to release in each of the next four months.
The Lake is currently 150 mm below flood stage. The dam is capable
of lowering the water level 200 mm each month, but additional
precipitation and evaporation affect the dam. The weather near Clear
Lake is highly variable. The Dam Commission has divided the months
into two two-month blocks of similar weather. The months within each
block have the same probabilities for weather, which are assumed
independent of one another. In each month of the first block, they
assign a probability of 1/2 to having a natural 100-mm increase in
water levels and probabilities of 1/4 to having a 50-mm decrease or a
250-mm increase in water levels. All these figures correspond to
natural changes in water level without dam releases. In each month
of the second block, they assign a probability of 1/2 to having a
natural 150-mm increase in water levels and probabilities of 1/4 to
having a 50-mm increase or a 350-mm increase in water levels. If a
flood occurs, then damage is assessed at $10,000 per mm above flood
level. A water level too low leads to costly importation of water.
These costs are $5000 per mm less than 250 mm below flood stage. The
commission first considers an overall goal of minimizing expected
costs. This model only considers this first objective.
Birge, R, and Louveaux, F V, Introduction to Stochastic Programming.
Springer, 1997.
$offtext
Sets p Precipitation levels in each month / low, normal, high /
t Time periods / dec, jan, feb, mar, apr /;
Table deltastoch(t,p) reservoir level change data for each season
low normal high
jan 50 150 350
feb 50 150 350
mar -50 100 250
apr -50 100 250;
Parameter
pr(p) Probability distribution / low 0.25, normal 0.50, high 0.25 /
floodCost 'K$/mm for amounts over flood level' / 10 /
lowCost 'K$/mm for amounts 250mm below flood level' / 5 /
linit initial water level /100/
delta(t) random changes in reservoir level for each season;
Variable
obj cost
l(t) level of water in dam end of period
r(t) mm released normally
f(t) mm of floodwater released
z(t) mm of water imported;
Positive variable l, r, f, z;
r.up(t) = 200;
* water level l is relative to 250mm below flood stage
l.up(t) = 250;
l.fx('dec') = linit;
delta(t) = sum(p, pr(p)*deltastoch(t,p));
Equations defobj, ldef(t);
defobj.. obj =e= sum(t$(ord(t)>1), floodCost * f(t) + lowCost * z(t));
ldef(t)$(ord(t)>1)..
l(t) =e= l(t-1) + delta(t) + z(t) - r(t) - f(t);
model mincost / defobj, ldef /;
file emp / '%emp.info%' /; emp.nd=2; put emp '* problem %gams.i%';
loop(t$(ord(t)>1),
put / 'stage ' ord(t):2:0 ' ' l.tn(t) ' ' r.tn(t) ' ' f.tn(t) ' ' z.tn(t) ' ' ldef.tn(t) ' ' delta.tn(t)
put / 'randvar ' delta.tn(t) ' discrete ';
loop(p, put pr(p) deltastoch(t,p)));
putclose;
Set s scenarios / s1*s81 /
Parameter
s_delta(s,t) random variable realization
s_l(s,t) level of water in dam end of period by scenario
s_r(s,t) mm released normally by scenario
s_f(s,t) mm of floodwater released by scenario
s_z(s,t) mm of water imported by scenario
s_obj(s) cost by scenario;
Set dict / s. scenario.''
delta.randvar. s_delta
l. level. s_l
r. level. s_r
f. level. s_f
z. level. s_z
obj. level. s_obj /;
solve mincost min obj using emp scenario dict;
```