Characterization and Estimation of Wind Energy Resources Using Autoregressive Modelling and Probability Density Functions

Abstract

The commonly used two-parameter Weibull distribution has fitted many wind speed distributions reasonably well. However, the use of such probability density functions (PDFs) does not provide a very accurate description of wind resources. The main drawback is that the wind time series data are assumed to be uncorrelated over time. This paper studies autoregressive (AR) modelling as an alternative method to characterize a given wind regime. The AR model allows for a very accurate representation of the wind resource and additionally, provides a tool for its forecasting. The AR model is compared with commonly used non-Gaussian distributions for two Caribbean wind-speed time-series. The comparison evaluates the fitting of the models, and, most importantly, the estimated values of power and energy to be extracted from the wind resource. It is concluded that autoregressive (AR) modelling provides greater accuracy than the popular Weibull and Rayleigh distributions, especially for low wind-speed systems.