Reducing Uncertainty in Wind Energy Resource Assessment by using Multivariable Distribution Model
Date Issued
2019-06
Author(s)
Demerdziev, Kiril
Abstract
Having reliable and precise wind energy resource assessment is essential for further analysis for conversion of wind energy into electricity. Previous practices that are common for representing wind data by sector-wise Weibull distribution, over time have been replaced with different multivariate and multimodal wind distribution models which are far more precise. The paper presents an upgraded model for accurate characterization and predict the annual variation of wind conditions. It is proven that the assumption of a constant air density value, can lead to notable differences between the predicted and real wind power available at a given site. Therefore, along with the main wind parameters, speed and direction, air density treated as a variable in this paper.
The method, based on the Multivariate Kernel Distribution model, is an improvement of the existing methods for representing the wind regimes. Before representing the multivariable wind distribution, a piecewise Bivariate probability density function is constructed. It is important to note that when modelling wind with sector-wise Weibull, we are assuming that the wind speed satisfies the same probability distribution inside a direction sector. Following, a Bivariate probability density function using piecewise joint distribution is carried out. Namely, this distribution contains all input parameters for calculating the multivariable Kernel distribution. For comparison of these two distributions, coefficients of determination are used. From Kernel's probability distribution function, the three parameters of the wind (speed, direction and air density) are further treated as continuous variables, which facilitates and refines all further steps for optimizing the distribution of wind turbines in one wind farm.
Aside of the other advantages, this model provides information on strengths of wind speeds and the energy content in them, it also enables selection of the appropriate type of wind turbine design for deployment at a given location.
The measured wind data used in this paper are from one existing wind farm and one wind measuring station with a good potential for further investigation. By this approach, we can calibrate model adequacy and ascertain a model that will serve as a referent guidance in the planning of future onshore wind farms.
The method, based on the Multivariate Kernel Distribution model, is an improvement of the existing methods for representing the wind regimes. Before representing the multivariable wind distribution, a piecewise Bivariate probability density function is constructed. It is important to note that when modelling wind with sector-wise Weibull, we are assuming that the wind speed satisfies the same probability distribution inside a direction sector. Following, a Bivariate probability density function using piecewise joint distribution is carried out. Namely, this distribution contains all input parameters for calculating the multivariable Kernel distribution. For comparison of these two distributions, coefficients of determination are used. From Kernel's probability distribution function, the three parameters of the wind (speed, direction and air density) are further treated as continuous variables, which facilitates and refines all further steps for optimizing the distribution of wind turbines in one wind farm.
Aside of the other advantages, this model provides information on strengths of wind speeds and the energy content in them, it also enables selection of the appropriate type of wind turbine design for deployment at a given location.
The measured wind data used in this paper are from one existing wind farm and one wind measuring station with a good potential for further investigation. By this approach, we can calibrate model adequacy and ascertain a model that will serve as a referent guidance in the planning of future onshore wind farms.
Subjects