Modelling wind behaviour is a complex problem. Knowledge of future wind speeds and directions could significantly help wind turbine designers, investors and operators, but prediction with a degree of practical accuracy is near impossible. Avraam Charakopoulos and colleagues at the University of Thessaly and the University of West Attica in Greece now study patterns in wind velocity and direction data using methods for complex systems analysis, paving the way for potentially better predictions in the future.
The researchers use speed and direction data collected at ten minute intervals from a wind turbine in Achaia, Greece, spanning March to December 2005. First they develop recurrence plots for velocity and angle of wind, which are visual representations of irregularities and periodicity in chaotic systems. The recurrence plot for wind velocity shows at what time a certain measured velocity of wind will recur on a two-dimensional plot of time against time. They find two time intervals, 2–4.5 days and 5–8.5 days, during which the variation of wind speed is correlated in a time series and lends itself better to predictive modelling. A network graph of the data showed clusters of similar observations. The number of similar observations or connections in these clusters distributed according to a power law, indicating that some wind speeds repeated more than others. This kind of pattern recognition could be used to identify where predictive models can be more accurate in their application.