Geoscience Reference
In-Depth Information
where
c
= crops, 1-6 (single crop, first crop, second crop paddy, groundnut,
sugarcane, and dry crops),
i
= nodes in Zones II-V or subarea,
B
i,c
= net annual benefit after production cost and water cost for crop
c
grown in subarea
i
in Rs/Mm
2
(Rs = Rupees, Indian Currency),
= land area under crop
c
in subarea
i
in Mm
2
.
A
i,c
The accounting system of water, which is still in use in Vaigai system,
is taken into consideration in the model with all other constraints as:
1. crop water requirement constraints,
2. land area constraints,
3. surface water availability constraints,
4. groundwater availability constraints,
5. continuity constraints.
The detailed linear programming model has been discussed in Ref. 6. It is
anoptimizationmodelwithin-depthdiscussiononmodelinginanarrayof
application areas using Simplex method. The model is run for 29 years of
inflow (1969-1997) into the basin.
Second, an artificial neural network (ANN) structure shown in Fig. 2
can be applied to develop an e
cient decision support tool considering
the parameters, which are non-linear in nature and to avoid addressing
the problem of spatial and temporal variations of input variables. By this
work an ecient mapping of non-linear relationship between inflow, storage,
demand and release pattern into an ANN model is performed for better pre-
diction on optimal releases from Vaigai dam. The back-propagation network
(BPN) is the most popular network among recent applications of ANN.
7
Feed forward Error Back-propagation Network (FFBPN) with Levenberg
Marguardt model is considered in this study. The hyperbolic tangent trans-
fer function, log sigmoid transfer function; the normalized cumulative delta
learning rule and the standard (quadratic) error function were used in the
framework of the model.
E
(
t
)=
1
2
(
d
j
(
t
)
y
j
(
t
))
2
,
−
(2)
where
E
(
t
) is the global error function at discrete time
t
and
y
j
(
t
)isthe
predicted network output at discrete time
t
and
d
j
(
t
) is the desired network
output at discrete time
t
.
