Agriculture Reference
In-Depth Information
way in which the birds should be fed to maxi-
mize or minimize a given objective function.
An optimization tool developed by EFG Soft-
ware (1995) combines three types of computer
program, namely, a feed formulation program
using linear programming, a broiler growth
model and an optimization algorithm. The
flow of information between these components
approximates a hill-climbing algorithm, which
is defined below.
Hill-climbing algorithms are commonly
used in optimization. These start with a given
solution and then proceed deterministically,
by following the direction of the gradient, to
the optimum solution. Such an algorithm is
likely to fail where there are local optima, since
it will stop at the first, and not necessarily the
highest, peak. Heuristic methods (Feigenbau-
em and Feldman, 1963) offer a wide range of
techniques to solve this problem. Many of the
heuristic techniques have been inspired by
evolutionary or biological processes, examples
being genetic algorithms and differential evo-
lution (Gilli and Winker, 2008). Allowing for
an impairment of the objective function, main-
taining a whole population of solutions at one
time, the deliberate use of randomness for cre-
ating new candidate solutions and local search-
ing are common principles used in heuristics.
In spite of the enormous increase in computing
power these techniques are still time consum-
ing, and do not offer a solution as rapidly as
would a least-cost feed formulation problem,
for example. However, the advantage of using
statistical and modelling techniques to optimize
performance rather than relying on the results
of a series of commercial trials compensates
fully for the delay.
A robust but simplified method that
works well is to explore the entire response
surface using a grid that could initially be
rough, and that is used simply to identify
the area in which the highest peak resides;
thereafter a hill-climbing algorithm is used
to home in on the optimum point. Such a
method is used in the EFG optimization pro-
cedure. The optimizer defines nutritional
constraints for practical broiler feeds. These
are passed to the feed formulation program
where the least-cost feed that meets these
constraints is determined. The characteris-
tics of this formulated feed are then passed,
as input, to the broiler growth model. The
performance expected from this feed when
given to a defined flock of broilers in a given
environment is predicted by the model, and
this predicted performance is then passed to
the optimizer to complete the cycle. The
next cycle starts with the optimizer modify-
ing the feed specifications, moving, accord-
ing to some in-built rules, to an optimum
point. The objective function to be opti-
mized can be defined in terms of any output
from the broiler growth model, but realistic-
ally would be an economic index of some
sort, such as margin over feed cost or margin
per m
2
/year.
The system is extremely versatile, allow-
ing for a wide range of management practices,
such as separate sex rearing, multiple harvest-
ing from one flock and sales of any mixture of
whole-bird, processed or further processed
products from which the revenue is derived.
Typical economic variables are included, al-
though these are readily customized to fit with
individual enterprises. The key to this ap-
proach lies in the ability of the broiler growth
model to reflect accurately the performance
expected under commercial conditions.
The simulation model, which predicts
food intake and hence growth rate and carcass
composition, is embedded in the Windows-
based feed formulation program, WinFeed.
Practical feed specifications are set up in
the usual way and these provide the starting
point for the optimization. A feeding pro-
gramme is set up from these formulated
feeds in which the amount, the form of the
feed (mash, crumbles, pellets), the propor-
tion of fines, and the cost of manufacture
and transport are defined. The optimization
process takes account of all the other set-
tings in both the feed formulation program
(feed prices, feed and nutrient constraints,
etc.) and broiler model (genotype, environ-
ment, fixed and variable costs, sources and
rates of revenue).
Optimizations Applied in EFG Broiler
and Pig Growth Models
At present the program optimizes three
aspects of a commercial broiler feeding
