Page 154 - IJOCTA-15-1

N. Tekbıyık-Ersoy / IJOCTA, Vol.15, No.1, pp.137-154 (2025)
5.3. Case 3 results than 10% implies an excellent forecasting. Hence,
it can be said that the forecasting performances
Case 3 includes four modeling parameters, in- of the developed models are excellent. As also
dicated with N:4 in Figure 4. The minimum
and mean MAE of 210 optimization simula- seen from the results, MAEOPT performed bet-
ter than MLR in almost all the cases, based on
tions/regression estimations are also shown on the
MAE and MAPE. Moreover, the forecasting per-
figure. Obviously, the average MAE dropped even
formances of the models developed for Case 3, are
further reaching around 1054.46 for MAEOPT.
superior to the ones in Case 1 and Case 2, in terms
C: 146 of 210 indicates that the minimum MAE
has been reached in 146 th combination. As seen of MAE and MAPE. That means although all six
rd
from the figure, combination of 3 , 4 th and 6 th models are successful, the best model to be used
and 7 th parameters resulted in the lowest MAE. for modeling and forecasting T¨urkiye’s renewable
energy capacity in the future is the one provided
This indicates that the model corresponding to
in equation (8) with the values in Table 7.
combination 146 for MAEOPT is the best model
among all the models developed in this paper (750 6. Conclusion
models), and it has an excellent performance. Re-
rd
ferring to Section 4, 3 , 4 th and 6 th and 7 th pa- Considering that the effects of global warming and
rameters are Total population, Urban population climate change have become more evident in the
(% of total population), Net energy imports (PJ), last decade, the importance of renewables has in-
and Coal imports (TJ) respectively. Considering creased significantly. Many countries around the
all the factors mentioned above, the best models world are investing in renewables, and trying to
for MLR and MAEOPT can be written as follows: integrate more renewable energy into their energy
mix, in order to reduce their carbon footprint. Es-
ˆ y i = b 1 x 1i + b 2 x 2i + b 3 x 3i + b 4 x 4i + b 0 (8)
pecially in the last two decades, T¨urkiye followed
where ˆy i represents the installed RE capacity in a similar trend and increased the amount and the
year i (in MW) and b 1 , b 2 , b 3 , b 4 , b 0 are as shown capacity of renewable energy installations in the
in Table 8. x 1i ,x 2i , x 3i and x 4i represent the Total country. This paper investigated the reasons be-
population, Urban population (% of total popula- hind this rapid development. The study focused
tion), Net energy imports (PJ), and Coal imports on three cases, and in each case, the number of
(TJ) in the i th year, respectively. modeling parameters have been increased. This
helped in understanding the importance and pri-
ority of the modeling parameters, mainly the most
5.4. Prediction performance of the
important factors affecting the renewable energy
developed models
development in T¨urkiye. The study performed in
Figures 5-7 show the prediction/estimation per- this paper revealed that the most important fac-
formances of the best models in each case. As tors affecting renewable energy installations are
seen from the figures, all six models are suc- the Total population and the Urban population
cessful in predicting the installed RE capacity in (% of the total population). The other important
T¨urkiye during the modeling period. After that, factors were recorded to be Net energy imports
some deviations are observed in terms of perfor- (PJ) and Coal imports (TJ) respectively. This
mance. In order to better compare the forecasting indicates that the energy import dependency in
performance of all these six models, their MAE, T¨urkiye has a potential impact in the future of
MAPE and RMSE values are also compared dur- RE development. The results also reveal that the
ing the whole prediction/forecasting period (in- need for importing coal (one of the main sources
cluding the 15 years modeling period and the 2 of electricity generation in T¨urkiye) also drives
years testing period). The results can be seen in the country’s installed RE capacity. It is clear
Tables 9-11. that T¨urkiye is trying to decrease its energy de-
pendency by utilizing more renewables. Hence, in
order to increase the utilization of RE in T¨urkiye
Based on the results provided in Tables 9-11, as and thus the energy dependency, the link between
expected, the MAE, MAPE and RMSE values of imports and renewable energy should be investi-
the best models found by MLR and MAEOPT gated in more detail.
have increased when the whole forecasting period The simulation results have shown that the per-
is considered (2005-2021). However, despite this formances of all the best models presented in this
increase, the MAPE values of all six models (one paper (6 models: 3 for MLR and 3 for MAEOPT)
for MLR and one for MAEOPT for each case) are excellent for the forecasting period, due to the
are less than 5%. According to, 30 MAPE of less MAPE’s being between %3.3107-%4.0963.
148