iSIGHT1. This involved linking the
(landside culvert). Particular empha-
optimum solutions throughout the
sis was the determination of appro-
parameter space. The numerical
numerical model of Lock 1 with
priate loss coefficient values for the
optimization techniques employed
iSIGHT. The optimization routine
intakes and the "chute" vertical tran-
were of the direct methods type.
was developed to automatically
sition. This transition involves two
Specifically, the method of feasible
change the specified coefficients in
vertical curves and a 9.1-m drop in
directions and modified method of
the model input file, execute the
culvert elevation immediately
feasible directions were used to find
LOCKSIM program, read the flow
upstream of the filling valves. The
local optima. Global optimization
solution, and compute error indica-
field data were used to establish the
was achieved using the explorative
tors. The error indicators were
values of these loss coefficients by
techniques of genetic algorithm and
chosen to be the differences in com-
integration of the numerical model
adaptive simulated annealing. Model
puted and observed operation time,
with the optimization software in the
simulation runs were completed in
pressure downstream of the valve
manner previously described. The
an automatic fashion for both the fill-
and the water surface in the valve
results of the adjusted model are
ing system and the emptying system
well at critical times during the oper-
shown on Figure 2. The model
in order to establish the loss
ation. The optimization scheme
reproduces the field data quite well
coefficients.
drove these error indicators toward
except for the pressures down-
zero by adjusting the specified
stream of the filling valve during the
energy loss coefficients. Techniques
Filling System
first 50 sec of the filling operation.
of both exploitative and exploratory
The valve opens about 55 percent
optimization were used. The explor-
Determination of the loss coeffi-
during this period. Early in the oper-
atory techniques, which are numeri-
cients for filling system components
ation, the computed pressures are
cal optimization techniques, provided
used the field data recorded for a
significantly higher than the mea-
minimization of an objective function,
single-valve filling operation
sured values due to the errors in the
while exploration was used to find
Figure 2. Model validation results for single-valve-filling operation, upper pool 221.4, lower pool 210.1
Engineous Software, Inc. No endorsement of this product is made or implied in this paper.
1
39
Previous Page
