A coupled system of stochastic differential equations for probabilistic wind speed modelling

Wind-based electricity generation plays a vital role in the renewable energy industry. Benefits from its use can only be obtained if it is possible to accommodate its variability and limited predictability. Stochastic differential equation (SDE) based approaches have demonstrated an improved capability of predicting temporal wind speed patterns that have statistical properties that are similar to those observed in reality. However, no standard approach for deriving such models exist due to the wide variations in the temporal statistical properties that one observes in wind speed data measured from location to location. Wind speed data which have been recorded at coastal locations, exhibit non-stationary features, like diurnal effect, seasonal effects, and temporal trends. In this work, such effects are eliminated using standard smoothing techniques.

A coupled system of second-order ordinary linear differential equations driven by a white noise forcing term was used for the probabilistic modelling of the residual data. The model was then used to predict the wind speed distribution and the corresponding autocorrelation function of wind speed data, recorded at the wind measurement centre of Kokkilai in northern Sri Lanka, from February 2015 to February 2016. Finally, the results of this novel approach were compared against the probabilistic modelling methods existing in the literature.