Page 156 - IJOCTA-15-1

N. Tekbıyık-Ersoy / IJOCTA, Vol.15, No.1, pp.137-154 (2025)

MLR (N: 4) / C: 146 of 210 / C : [3 4 6 7]
B
MAEOPT (N: 4) / C: 146 of 210 / C : [3 4 6 7]
B
55000
Real RE
50000 Predicted RE (MLR)
Predicted RE (MAEOPT)
45000
Installed RE Capacity (MW) 35000
40000

30000
25000
20000

15000
10000
2005 2010 2015 2020

Figure 7. Prediction Results for Case 3

Table 9. MAE Comparison of MLR and MAEOPT for Forecasting

Criteria MLR MAEOPT % Reduction
Case 1 1077.1727 1016.9075 5.59
Case 2 1052.2955 1055.6287 -0.31
Case 3 983.5615 939.0191 4.52

Table 10. MAPE Comparison of MLR and MAEOPT for Forecasting

Criteria MLR MAEOPT % Reduction
Case 1 4.0963 3.8827 5.21
Case 2 3.9489 3.7583 4.82
Case 3 3.6954 3.3107 10.40

Table 11. RMSE Comparison of MLR and MAEOPT for Forecasting

Criteria MLR MAEOPT % Reduction
Case 1 1474.1243 1426.6729 3.21
Case 2 1622.4301 1749.6430 -7.84
Case 3 1516.4849 1595.8979 -5.23

The results also revealed that the best model de- Appendix A. Analytical proof for
veloped for Case 3 (MAEOPT), (considering all optimal MLR results
the four modeling parameters; Total population,
Urban population (% of total population), Net This section will provide the derivation of the an-
energy imports (PJ), and Coal imports (TJ)), alytical equations that need to be solved in order
to find the optimum solution for MLR. MLR is
can safely be used for predicting the future of
based on the minimization of sum of the squared
renewable energy development in T¨urkiye. It
errors (SSE). Hence, in order to find the optimum
should also be noted that, in the absence of data
values of b 0 and b k , the derivative of SSE with re-
availability, the MLR and MAEOPT models de-
spect to b 0 and b k should be found and should be
veloped in Case 1 and Case 2 can also be used for
equated to 0. Solving those equations would yield
prediction in order to provide rough estimates or
the optimum values of b 0 and b k . The following
benchmark for the future.
set of equations show the related derivation.

150