Stochastic modelling of wind velocities time series

Abstract

The knowledge of wind speed time series is necessary to evaluate, for instance, the power values produced by wind turbines, in order to investigate load matching and storage requirements. The usual description of the stochastic properties of wind speed independent of its direction is here criticized and an approach based on the coupled analysis of both wind speed and direction is proposed. Two stochastic models simulating three hourly wind speed and direction time series are investigated; encouraging results are obtained with a scheme combining a Markov chain model and a Discrete Auto Regressive model of order 2 (DAR(2)). A comparison with 42 years of wind data measured at Brindisi airport, Italy, is made, showing that this compound model is able to generate direction and speed sequences that fairly agree, in the statistical sense, with the observed time series at the considered site.