A combined simulation–GA optimization model has
been developed in this study to determine the optimal
upper and lower rule curves of a reservoir. The model
is applied to the reservoir of the Nam Oon Irrigation
Project in Thailand to determine the operational rule
curves and at the same time compute storage volumes
and reservoir water releases. It is found that the combined
simulation–GA approach provides robust and acceptable
solutions in determining the optimal upper and lower
rule curves. In this approach, the objective function is
optimized based on maximum net benefit with penalty
function. There are three sensitive parameters, namely
population size, crossover and mutation probabilities,
which are adjusted to determine the best fitness or optimal
solution. The optimal upper and lower rule curves and the
maximum net benefit are determined by using monthly
data from 1976 to 1998. For scenario 1, the maximum net
benefit obtained by the combined simulation–GA, HEC-
3 and SOP methods are nearly the same. However, in
scenarios 2 and 3, in which cropping areas are different
from scenario 1, the percentage losses from the combined
simulation–GA method are less than that determined by
the SOP method. Another point is that the combined
simulation–GA method requires less user experience and
takes less time in data preparation and computation than
the HEC-3 method. The model software for the combined
simulation–GA method developed in this study is user
friendly with a graphical interface for data input and
output presentation.
From the foregoing sensitivity analysis and the results
of proportion of maximum fitness, it is found that the
GA is robust and provides the best performance over a
wide range of population size, crossover and mutation
probabilities. The effects of population size, crossover
and mutation probabilities are determined by varying
them one at time. In this case study, it is found that
the population size, crossover and mutation probability
have only a slight influence on the performance of the
GA. The most suitable values of these parameters in