Beta-Generalized Lindley Distribution: A Novel Probability Model for Wind Speed
Tiantian Yang and Dongwei Chen · 2025
Read Beta-Generalized Lindley Distribution: A Novel Probability Model for Wind Speed by Tiantian Yang and Dongwei Chen (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-09912.txt.
Introduction
Tiantian Yang and Dongwei Chen introduce a new probability model for wind speed in this 2025 paper. Fitting wind data well matters for energy forecasts and safe building design, above all in the heavy right tail of rare gusts. Their model, the beta-generalized Lindley distribution, has four adjustable parameters. They fit it to a decade of tower readings taken at four heights in Colorado.
Across density fit and standard goodness checks, the new model beat its own simpler cousins. It also beat common rivals like Weibull and log-normal fits. The edge was clearest at the ninety-fifth and ninety-ninth percentile gusts, the values engineers care about most. The authors present the model as a strong new option for wind studies.
Catalog record
| CREATOR | Tiantian Yang and Dongwei Chen |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-09912
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Beta-Generalized Lindley Distribution: A Novel Probability Model for Wind Speed Thanks: Department of Mathematics and Statistical Sciences, University of Idaho, Moscow, ID, USA, 83843; Email: [email protected] Thanks: Department of Mathematics, Colorado State University, Fort Collins, CO, US, 80523; Email: [email protected] Thanks: Corresponding author: Dongwei Chen. Email address: [email protected] Tiantian Yang and Dongwei Chen Abstract . Wind speed distribution has many applications, such as the assessment of wind energy and building design. Applying an appropriate statistical distribution to fit the wind speed data, especially on its heavy right tail, is of great interest. In this study, we introduce a novel four-parameter class of generalized Lindley distribution, called the beta-generalized Lindley (BGL) distribution, to fit the wind speed data, which are derived from the annual and long-term measurements of the Flatirons M2 meteorological tower from the years 2010 to 2020 at heights of 10, 20, 50, and 80 meters. In terms of the density fit and various goodness-of-fit metrics, the BGL model outperforms its submodels (beta-Lindley, generalized Lindley, and Lindley) and other reference distributions, such as gamma, beta-Weibull, Weibull, beta-exponential, and Log-Normal. Furthermore, the BGL distribution is more accurate at modeling the long right tail of wind speed, including the 95 t h 95^{th} and 99 t h 99^{th} percentiles and Anderson–Darling statistics at different heights. Therefore, we conclude that the BGL distribution is a strong alternative model for the wind speed distribution. 2020 Mathematics Subject Classification 62P12 1. Introduction Better knowledge of wind speed distribution is of high interest for many applications. For example, solid wind speed analysis can improve air transport safety since high wind speed may cause turbulence and associated problems at taking off and landing [ 22 ] . A good estimation of the wind speed distribution can also contribute to risk assessment of insurance [ 15 ] , pollutant dispersion control in urban areas [ 24 ] , and agricultural management [ 5 ] . Furthermore, it contributes to wind engineering such as building and bridge design [ 3 ] , since defining the design wind speed is a key step in evaluating the wind loading of structures and their safety under the wind [ 26 ] . In particular, accurate estimation of wind speed distribution is critical to the assessment of wind energy potential, the site selection of wind farms, and the operation management of wind power conversion systems [ 17 , 23 ] , since researchers often use the wind speed distribution to estimate the wind energy potential and assess the average annual wind energy yield of a wind turbine at a certain location before its installation [ 10 ] . Therefore, it is necessary to estimate the wind speed distribution accurately. The statistical distribution of wind speed has its characteristics: the right tail is heavy. Many probability distributions have been applied to fit the wind speed data, such as the two-parameter Log-Normal distribution [ 7 ] , two-parameter Weibull distribution [ 4 ] , two-parameter generalized Lindley distribution [ 2 ] , three-parameter extended generalized Lindley distribution [ 13 ] , upper-truncated Weibull distribution [ 12 ] , new alpha power transformed power Lindley distribution [ 1 ] , two-parameter gamma distribution and one-parameter Rayleigh distribution [ 25 ] . Among the distributions used to fit the wind speed, the top four most used distributions are the two-parameter Weibull (W), two-parameter Log-Normal (LogN), two-parameter gamma (GAM), and one-parameter Rayleigh (R) distributions [ 11 ] . And by far, the most frequently used statistical distribution is the two-parameter Weibull distribution [ 11 ] , denoted as W(\alpha,\lambda) , where its shape ( \alpha ) and scale ( \lambda ) parameters vary with location and season [ 24 ] . Note that the one-parameter Rayleigh distribution is a special case of Weibull distribution when \alpha = 2. Interested readers may refer to [ 25 , 11 ] for more information about a summary of the probability distributions used for wind speed. In this work, we introduce a novel four-parameter statistical model for wind speed, which outperforms other reference distributions and is known as the beta-generalized Lindley (BGL) distribution [ 21 ] . The BGL model has four parameters (\alpha,\lambda,a,b) and is often used to fit data from a long-tail distribution. Its probability density function (PDF) is \begin{split}f_{BGL}(x;\alpha,\lambda,a,b)=&\frac{\alpha\lambda^{2}(1+x)e^{-\lambda x}}{B(a,b)(1+\lambda)}\left[1-\frac{1+\lambda+\lambda x}{1+\lambda}e^{-\lambda x}\right]^{a\alpha-1}\\ &\times\left\{1-\left[1-\frac{1+\lambda+\lambda x}{1+\lambda}e^{-\lambda x}\right]^{\alpha}\right\}^{b-1},\end{split} where x\geq 0 , \alpha>0 , \lambda>0 , a>0 , and b>0 ; refer to Section 3 in this paper for the details of the BGL density and other statistical properties, or the original BGL paper [ 21 ] . The four-parameter BGL distribution is proven to have good statistical properties and can capture different shapes of the density function, especially for the long tail [ 21 ] . Hence, it is suitable to fit the wind speed data. This paper is organized as follows. In Section 2 , we introduce the background of the used wind speed data and its data preprocessing procedure. Section 3 introduces the statistical properties of beta-generalized Lindley (BGL) distribution, other reference distributions, and multiple goodness-of-fit evaluation metrics used in this study. In Section 4 , we compare the performance of BGL distribution with other reference distributions on the long-term wind speed data from the years 2010 to 2020 (2010-2020) at four heights (10, 20, 50, 80 meters) using different evaluation criteria; we also investigate the performance for selected annual wind speed in the years 2010, 2015, 2020 at height 80 meters (m). The results show that the BGL distribution performs better, especially on the right tail, including the 95^{th} and 99^{th} percentiles. Finally, we conclude our results and propose future work in Section 5 . 2. Data Background and Preprocessing In this study, we use the wind speed data (m/s) derived from the measurements of the Flatirons M2 meteorological tower [ 9 ] , located about 8 km south of Boulder, Colorado, USA. The tower is about 82 meters tall and located at 39.9106° N and 105.2347° W, with its base at 1855 m above the mean sea level. We use hourly averaged wind speed data from 00:00:00 of January 1st, 2010, to 23:00:00 of December 31st, 2020, measured at four heights of the M2 tower, ranging from 10, 20, 50, and 80 m above the ground, respectively. The original wind speed data can be downloaded at [ 8 ] . We conduct data preprocessing as some timestamps in the original data are repeated, and the corresponding wind speeds are marked at -99999 to show the abnormality. However, there were only fourteen repetitions during the period used, and we deleted these repetitions. The total number of observations for each height for the long-term wind speed data for the years 2010 to 2020 (2010-2020) is 96,432. The descriptive statistics of the wind speed data (m/s) at heights 10, 20, 50, and 80 m for the years 2010-2020 are listed in the Panel A of Table 1 , including the sample size ( n ), minimum (min), maximum (max), median, mean, variance, skewness, and kurtosis. The sample skewness and kurtosis are calculated by \text{skewness}=\frac{\frac{1}{n}\sum_{i=1}^{n}(x_{i}-\bar{x})^{3}}{\left[\frac{1}{n}\sum_{i=1}^{n}(x_{i}-\bar{x})^{2}\right]^{3/2}},\ \text{and}\ \text{kurtosis}=\frac{\frac{1}{n}\sum_{i=1}^{n}(x_{i}-\bar{x})^{4}}{\left[\frac{1}{n}\sum_{i=1}^{n}(x_{i}-\bar{x})^{2}\right]^{2}}. The skewness is zero for data sampled from the normal distribution, while the kurtosis is three. A positive skewness indicates the