As can be observed in Fig. 5, except for GDP which has been
growing smoothly since 1980, all independent parameters have
been fluctuating e particularly since 2005. Because of possibility of
non-linear correlation between the independent parameters and
the energy demand, the neural network modeling has been chosen
for this study. A non-linear transfer function is embedded in this
ANN to grasp this non-linear relation of the parameters. In order to
estimate the future trend of energy demand in the industrial sector
based on independent parameters, we need to anticipate the
behavior of independent parameters in the future. For the future trend of GDP, a second-order polynomial equation was fit to the
GDP growth curve. The value of R2 shown in the GDP graph of Fig. 5
confirms the accuracy of the fitted curve. For the other independent
variables, we define a scenario which is called the CPS (Constant
Price Scenario). In this scenario the price remains at the level of the
average price of last five years of data set.