Inverted Kumarswamy distribution for modeling the wind speed data: Lake Van, Turkey
Introduction
The world's energy consumption is increasing day by day. Moreover, limited and unsustainable fossil fuels mostly supply energy needs. It is a known fact that the use of fossil fuels has adverse effects on the environment. Also, the use of fossil fuels is not sustainable, both environmentally and economically. Consequently, switching to alternative renewable energy sources can be considered as a necessity.
Renewable energy sources play an important role among all energy sources as they provide sustainable economic development by taking environmental factors into account [1]. Wind energy is a renewable energy source that contributes not only to satisfy the increasing energy demand but also to reduce the effects of climate change.
Wind power generation has snowballed over the past two decades. It has been used as an essential renewable energy source in electricity production in many developed countries [2]. According to the 2017 report of Global Wind Energy Council (GWEC), global wind power market remained above 50 GW (GW) in that year. Nine states have more than 10,000 MW (MW) of installed capacity, including China, the United States of America (USA), Germany, India, Spain, the United Kingdom (UK), France, Brazil, and Canada. The wind power installations continue to spread in the rest of the world. Turkey has also been increasing emphasis on renewable energy sources see; Fig. 1. In Turkey, there are currently more than 30 projects amounting to 800 MW, under construction. The Ministry of Energy and Natural Resources also announced the national target for wind power is set at 20 GW by 2023 [3]. However, many regions of Turkey have not been examined yet.
According to the 2017 report of the GWEC, the data concerning global installed wind power capacities for the top 10 countries are illustrated in Fig. 1.
Some crucial points such as characteristics of the wind speed regimes and wind capacity of the candidate region should be considered in planning the installation of wind power plants and using wind turbines effectively [4]. Accurate determination of wind energy potential requires detailed information about the wind characteristics of the candidate region [5]. A small error in wind speed modeling may result in a more significant error in calculating the energy output [6].
In wind energy studies, the Weibull distribution introduced by Walodi Weibull in 1951 [7] having a probability density function (pdf) with a scale parameter σ and a shape parameter αis the most widely used distribution.
As it is known, skewness and kurtosis measures are functions of the shape parameter(s). If a distribution has only one shape parameter, skewness and kurtosis measures take values in a limited range for different values of the shape parameter. This result delimits the modeling capability of a distribution; see Fig. 4. In this context, the Weibull distribution has a deficiency in terms of the skewness and kurtosis measures, since it has only one shape parameter; see Rouse [8] for skewness and kurtosis measures of the Weibull distribution. Hence, some papers in the literature mentioned that the Weibull distribution might not adequately model all kinds of wind regimes. For example, Wais [4], Akgül et al. [6], Arslan et al. [9], and references therein stated that the Weibull distribution performs inadequate modeling for wind speed data having more skewness and/or kurtosis. Also, Takle and Brown [10] showed that problems may occur when using the Weibull distribution to model data including calm or very low wind speeds. Therefore, Takle and Brown [10] proposed to use hybrid density function in such a situation. Based on these conclusions, different distributions should be considered to model wind speed more efficiently and minimize errors in the wind power estimation.
The originality of this work comes from the fact that the three-parameter Inverted Kumaraswamy (IKum) distribution introduced by Al-Fattah et al. [11] is utilized to model wind speed data as an alternative to the Weibull distribution. The IKum distribution is obtained by transformation to the random variable T that follows the Kumaraswamy distribution. The pdf and cdf of the IKum distribution areandrespectively. Here σ is the scale parameter, and α and β are the shape parameters.
The IKum distribution exhibits a longer right tail than several widely-used distributions, and this feature positively affects the distribution's ability to fit the rare events occurring in the right tail, i.e., outlying observations at the right tail. Besides, it is a flexible distribution with two shape parameters along with a scale parameter. Therefore, values of skewness and kurtosis measures of the IKum distribution take values on a surface rather than a line as the Weibull distribution; see Fig. 4. The skewness and kurtosis measures of the IKum distribution have a more extensive range than the Weibull distribution. This feature makes it attractive in modeling the data having different skewness and kurtosis values. It should also be noted that the Lomax (Pareto type-II), Beta type-II (inverted Beta), and Log-logistic (Fisk) distributions are sub-models of the IKum distribution. Several distributions such as the exponentiated Weibull, exponentiated Burr type-XII, Kumaraswamy Dagum, Kumaraswamy inverse Weibull distributions, and sub-models of them can be obtained by applying appropriate variable transformation on the IKum random variable. Details of the IKum distribution are not provided here for brevity; see Al-Fattah et al. [11] for further information. The Weibull distribution is also taken into consideration due to its common usage to make the study complete.
This study has the following contributions to the related literature. (i) The suggested distribution, the IKum, is being used for the first time in modeling the wind speed data. (ii) The parameters of the IKum distribution are estimated by the maximum likelihood (ML), least squares (LS), and maximum product of spacing (MPS) methods. The ML, LS, and MPS methods' efficiencies are compared for the IKum distribution via a Monte-Carlo (MC) simulation study. To the best of authors’ knowledge, there is no study concerning the other estimation methods except the ML and Bayesian in the related literature. (iii) Since there have been plans for establishing a wind power plant to Van, this is the first comprehensive study conducted in the considered region. This region is located in the highlands of eastern Anatolia, Turkey, at , . It includes the largest soda lake on Earth.
The wind speed data used in this study are taken from the Turkish State Meteorological Service and contains wind speeds obtained from 6 meteorological stations Van, Gevas, Tatvan, Ahlat, Ercis, and Muradiye around Lake Van in Turkey. Gevas, Ercis, and Muradiye are the districts of Van province. Tatvan and Ahlat are the districts of Bitlis province. Bitlis has tourism potential with its natural beauties, e.g., Mount Nemrut stratovolcano with Nemrut Crater Lake and Mount Suphan stratovolcano. Also, Van has the largest population with over one million in eastern Anatolia, Turkey. It has recently been announced that 50 textile factories will be established in Van to improve its industrialization.
Renewable energy sources play an essential role in serving natural beauties and sustainable industrialization. The accurate estimation of the wind energy potential in this region will be essential for future investments to be efficient. In this context, the outcomes of this study will guide the decision-makers during the installation of wind power plants in Bitlis and Van.
Section snippets
Overview of the previous studies
In this section, some of the studies that have been done in the last decade are reviewed. Also, the distributions used in these studies are tabulated at the end of this section to make literature review more compact and easy to follow; see Table 1.
Several distributions such as the inverse Weibull (IW), Log-normal, Gamma, generalized Lindley (GL), and Kappa are used for modeling wind speeds in the literature. It has been seen that, in particular cases, these distributions modeling performance is
Wind speed data
The data sets analyzed in this study are taken from the Turkish State Meteorological Service. The wind speed data include average hourly measurements from September 2017 to September 2018 at 10 m height in districts located on the side of Lake Van; see Fig. 2 and Table 2 for the geographical information of each station. Note that Fig. 2 was created through Mapcreator [40], a website that allows customization of a map online, and the data provided in Table 2 are taken from Ref. [41].
Descriptive
Estimation methods
Al-Fattah et al. [11] used the ML and Bayesian methods in estimating the parameters of the IKum distribution. In this study, the ML, MPS, and LS methods are used for estimating the parameters of the IKum distribution. However, the ML, MPS, and LS estimators of the unknown parameters of the IKum distributions cannot be obtained in closed forms. Since corresponding equations for the parameters of interests involve nonlinear functions of unknown parameters; see also Bagci and Arslan [43] in the
Model evaluation and results
In this section, the criteria used to compare the distributions fitting performances are given. Then, the actual wind speeds are modeled by the Weibull and IKum distributions.
Conclusion
The outcomes of this study show that the IKum distribution is preferable over to the well-accepted Weibull distribution in terms of the RMSE, , and KS criteria for four of six stations. Note that the PDE criterion plays a crucial role in determining the wind energy potential of a region. In this study, more importantly, the IKum distribution also provides better modeling performances than the Weibull distribution in terms of the PDE criterion in five out of six stations; see Table 8.
The IKum
Credit author statement
Kubra Bagci: Conceptualization, Methodology, Software, Formal analysis, Writing. Talha Arslan: Conceptualization, Methodology, Software, Formal analysis, Writing-Review & Editing. H. Eray Celik: Data curation, Investigation.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
The authors are grateful to the editor, associate editor, and the anonymous reviewers for their valuable comments and suggestions, which substantially improved the paper.
References (43)
- et al.
Investigation of wind characteristics and assessment of wind energy potential for Waterloo region, Canada
Energy Convers Manag
(2005) - et al.
Monthly and seasonal assessment of wind energy characteristics at four monitored locations in Liguria region (Italy)
Renew Sustain Energy Rev
(2010) Two and three-parameter Weibull distribution in available wind power analysis
Renew Energy
(2017)- et al.
An alternative distribution to Weibull for modeling the wind speed data: inverse Weibull distribution
Energy Convers Manag
(2016) - et al.
Generalized Lindley and Power Lindley distributions for modeling the wind speed data
Energy Convers Manag
(2017) Modeling wind speed and wind power distributions in Rwanda
Renew Sustain Energy Rev
(2011)- et al.
Analysis of some flexible families of distributions for estimation of wind speed distributions
Appl Energy
(2012) Use of multi-parameter distributions for offshore wind speed modeling: the Johnson SB distribution
Appl Energy
(2013)- et al.
Multi-peak Gaussian fit applicability to wind speed distribution
Renew Sustain Energy Rev
(2014) - et al.
Analysis of the upper-truncated Weibull distribution for wind speed
Energy Convers Manag
(2015)
Evaluating the suitability of wind speed probability distribution models: a case of study of east and southeast parts of Iran
Energy Convers Manag
Global comparison of the goodness-of-fit of wind speed distributions
Energy Convers Manag
Deep assessment of wind speed distribution models: a case study of four sites in Algeria
Energy Convers Manag
Wind speed analysis using the extended generalized Lindley distribution
Renew Energy
Use of Birnbaum-Saunders distribution for estimating wind speed and wind power probability distributions: a review
Energy Convers Manag
Introducing a system of wind speed distributions for modeling properties of wind speed regimes around the world
Energy Convers Manag
Accurate estimation of T year extreme wind speeds by considering different model selection criterions and different parameter estimation methods
Energy
Comparative study of numerical methods for determining Weibull parameters for wind energy potential
Renew Sustain Energy Rev
Wind speed probability distribution estimation and wind energy assessment
Renew Sustain Energy Rev
A new estimation approach based on moments for estimating Weibull parameters in wind power applications
Energy Convers Manag
Comparison of seven methods for determining the optimal statistical distribution parameters: a case study of wind energy assessment in the large-scale wind farms of China
Energy
Cited by (23)
A new probabilistic model with applications to the wind speed energy data sets
2024, Alexandria Engineering JournalBrazilian wind energy generation potential using mixtures of Weibull distributions
2024, Renewable and Sustainable Energy ReviewsEconomic viability assessments of high voltage direct current for wind energy systems
2023, Sustainable Energy Technologies and AssessmentsA new generalization of Lindley distribution for modeling of wind speed data
2022, Energy ReportsCitation Excerpt :An et al. (2020) introduced a new probability distribution named “Alpha logarithmic transformed Log-normal distribution” and used it for modeling wind speed analysis. Bagci et al. (2021) utilized Inverted Kumaraswamy distribution to analyze the wind speed data recorded in Turkey. Neshat et al. (2021) made a significant contribution by attempting to increase the accuracy of wind speed prediction using a novel hybrid deep learning-based evolutionary strategy.