Data set on wind speed, wind direction and wind probability distributions in Puerto Bolivar - Colombia

This paper presents wind speed and direction data measured with a weather station located in Puerto Bolivar, department of La Guajira, situated in the extreme north of Colombia, whose geographic coordinates are 12°11′N 71°55′W. A wind speed and direction sensor, a barometric pressure sensor, and a temperature sensor were used to obtain the presented data. These data were taken at the height of 10 m, which is the highest point of the weather station. The data taken by the meteorological station correspond to a period of 20 years (1993–2013), with hourly frequency. For the missing data, a mathematical model to estimate the Julian averages was developed, allowing to calculate the frequency histograms and four types of probability distributions for these data. Also, the representative wind roses were generated, taking into account the averages in each of the 12 months of the year.


a b s t r a c t
This paper presents wind speed and direction data measured with a weather station located in Puerto Bolivar, department of La Guajira, situated in the extreme north of Colombia, whose geographic coordinates are 12 11 0 N 71 55 0 W. A wind speed and direction sensor, a barometric pressure sensor, and a temperature sensor were used to obtain the presented data. These data were taken at the height of 10 m, which is the highest point of the weather station. The data taken by the meteorological station correspond to a period of 20 years (1993e2013), with hourly frequency. For the missing data, a mathematical model to estimate the Julian averages was developed, allowing to calculate the frequency histograms and four types of probability distributions for these data. Also, the representative wind roses were generated, taking into account the averages in each of the 12 months of the year.

Data
The data presented in this paper are meteorological measurements taken at a weather station located in Puerto Bolivar, Colombia. Daily averages of wind speed and direction were obtained. The historical series has some missing data for the measured period. For this reason, in order to complete these data, it was necessary to use an algorithm developed in MATLAB® to calculate the Julian averages for each of the series. This average estimation uses wind speed and wind direction data from the same month, day, and time from other years to calculate the average of these and estimate the missing data.
The data series are presented, complemented by different probability distributions that describe the statistical behavior of the data which parameter are shown in Table 1, together with the superposition of four probability distributions obtained from the monthly wind speed and direction data supplied with this document. These distributions are shown in Fig. 1 and 2. In addition, the monthly wind rose graphs are presented in Fig. 3 and 4 to determine the most likely wind direction in the place studied. The original hourly data with which these roses have been generated are presented in the attached documents. The calculations to generate all these figures and tables were made with the data presented in Appendix A.

Experiment set up
The weather station was located in Puerto Bolivar, in the department of La Guajira. This department is located in the north extreme of Colombia at coordinates 12 11 0 N 71 55 0 W. In this place, was installed a wind speed sensor, wind speed sensor, and pressure sensor, and the parameters of the installed sensors are shown in Table 2.
The wind speed and direction were measured at the height of 10 m, this being the highest point of the weather station. A little further down, the temperature and relative humidity sensors were located. The schematic diagram of the weather station used to measure the data presented in Appendix A, is shown in Fig. 5 Value of the data The data provided in this work can be inputs for a projection of future behavior for this region of the country. The raw data supplied can be used to apply different wind speed and direction calculation procedures to them in future investigations.
The data from the wind direction and speed graph allow estimates to be made of the energy efficiency of wind turbines that can be used in this region.

Method
The data taken by the weather station correspond to 20 years (1993e2013), with hourly frequency. For the missing data, a data mining was performed to calculate the empty spaces and fill them with the averages from the data on the same day and time in different years, which are presented in Appendix A. Thus, from the totality of these data, four types of probability distributions are calculated. Therefore, the wind speed data are presented as a continuous random variable with an associated probability distribution, and the wind roses give the predominate wind direction presented every month. The methodological procedure developed for the data treatment is shown in Fig. 6, which consists of four fundamental stages.
In the first stage, the series of data from the sensors installed at the measurement site was read, followed by the development of software for mathematical data processing, allowing the user to select the variables and desired time interval for analysis. Once the variables have been selected, the statistical treatment of the data is performed to calculate the desired outputs, such as frequency histograms, wind roses, and wind speed probability distributions. Fig. 7 shows the main views of the WindAnalysisUA v1.0 software developed to analyze the wind data.
The four probability distributions used were Normal or Gaussian, Gamma, Weibull and Rayleigh [1], which are presented below along with their mathematical models.

Gaussian probability distribution
This distribution can be applied in a large number of case studies, and this makes it a distribution of great statistical relevance [2]. Therefore, this distribution adjusts greatly to physical measurements. This distribution is governed by the probability density function for a normally distributed random variable given by Equation (1) [3,4].

Gamma probability distribution
This model is commonly used to adjust wind speed distributions. Contrary to the symmetry presented by the Gaussian distribution, the gamma distribution is biased to the right. This function is given by GðaÞ ¼ R ∞ 0 x aÀ1 e Àx dx for a > 0 and the probability density function is calculated using Equation (2) [5]. where a is the form parameter, and b the scale parameter. When a; b > 0, the value of EðXÞ ¼ aq; and Var ðXÞ ¼ aq 2 .

Weilbull probability distribution
This is another widely applied model for the amplitude distribution of wind speeds over time. This distribution is influenced by the shape parameter (k or a), which varies between values 1 and 3.6. In addition, it depends on a scale parameter (c, q or b). If a random variable X fits the probability density function expressed by Equation (3), it can be said to have a Weibull distribution [6]. f ðx; a; qÞ ¼