**5. Results**

As one of the goals of the study is fitting WS data to TPWD and estimating parameters of the distribution, an R package called "DiceOptim" was used for applying EGO, a "DEoptim" package was used for applying DE, a "GA" package was used for applying GA, and an "optimization" package was used for applying SA [83–87].

Table 4 shows the estimated value of shape (*k*) and scale (*c*) parameters of the TPWD using four different techniques. From the table, it can be concluded that there are no huge differences between the parameters estimated using the four different techniques. Figures 8 and 9 represent the histogram of observed wind speed and the estimated TPWD obtained using four techniques per each month of the year. Table 5 and Figures 10 and 11 show the performance of techniques based on two different metrics: RMSE and R2. From Table 5 and Figures 10 and 11, it can be seen that EGO performs better than other techniques for estimating the parameters of TWPD; the larger the value of R<sup>2</sup> the better the performance of estimation as seen in Figure 10 that R<sup>2</sup> calculated for estimations.

**Figure 8.** Monthly probability density function estimation and monthly histograms of observed WS data (from January to June) (own elaboration).

**Figure 9.** Monthly probability density function estimation and monthly histograms of observed WS data (from July to December) (own elaboration).


**Table 4.** Estimated TPWD parameters for monthly wind speed data.

**Table 5.** Performance comparison based on different metrics.

