Modelling wind speed parameters for computer generation of wind speed in Flanders: a case study using small wind turbines in an urban environment

Since September 2010 wind speed measurements are conducted using an ultrasonic Doppler anemometer. The anemometer is mounted on a building in mid Flanders (Belgium) at a height of 12.3 m and takes samples at a frequency of 4 Hz. Using these 3D wind speed vector measurements, a potential for small scale wind turbines is estimated. Parameters of the wind are analyzed as for example: the frequency distribution of wind speed samples at different time averages, frequency distribution of wind direction, standard deviation of the wind speed and wind direction. Afterwards the basic properties, necessary for computer generation of wind data, are implemented in Matlab®. The goal of the paper is to reproduce wind speed and wind direction samples that model the wind in a realistic manner and with conservation of the observed characteristics of the real dataset. One way of implementing computer generated wind data is by the use of a Markovian model in combination with a probability integral transform. This way will preserve some properties of the wind data, but other properties are not preserved as for example the diurnal cycle of the wind data [1]. There are other ways to model wind data as for example by the use of an autoregressive process model based on the wind speed measurements [2]. Another way of generating computer wind data is by the use of the probability distribution of the wind data, which is widely accepted to be a Weibull distribution [3, 4] which can be a Raleigh distribution under certain conditions. To start this paper, a description is made on how these data are obtained by the anemometer. Some differences in calculating the time average of the wind speed are for example the difference between the time average of the wind vector and the time average of the wind scalar [5]. The relative difference between these quantities should be bigger when wind speeds are low, which happens to be the case in mid Flanders on moderate height. Afterwards different Markov chains are made based on the transition probability of the wind measurements in different seasons. Finally the computer generated wind speed time series are compared to the real measurements in order to evaluate the correctness of the computer generation model. If possible, other wind reproduction techniques will be compared to the Markovian model. [1] A. Shamad, M. A. Bawadi, W. M. A. Wan Hussin, T.A. Majid, S. A. M. Samussi, First and second order Markov chain models for synthetic generation of wind speed time series. Energy. 30, 693-708 (2005) [2] P. Poggi, M. Muselli, G. Notton, C. Cristofari, A. Louche, Forecasting and simulating wind speed in Corsica by using an autoregressive model. Energy Conversion and Management. 44, 3177-3196 (2003) [3] F. Youcef Ettoumi, H. Sauvageot, A. -E. -H. Adane, Statistical bivariate modeling of wind using first order Markov chain and Weibull distribution. Renewable Energy. 28, 1787-1802 (2003) [4] T. Ackermann, Wind Power in Power Systems (John Wiley & Sons, Ltd, England, 2005), pp. 452 [5]D.B. Gilhousen, A field evaluation of NDBC moored buoy winds. Journal of Atmospheric and Oceanic Technology. 4, 94-104 (1987)

ISBN:978-3-98 13870-5-

Jaar van publicatie:2012

Toegankelijkheid:Open