Predicting Operating Conditions in Wind Turbines Under Gusty Winds

This paper presents a comparison of four techniques for predicting future operating conditions. These methods were used to analyze the vibration data of a wind turbine that operates under gusty winds. The data were recorded at a 12 m diameter HAWT gearbox, and they were preprocessed to select the most significant frequency. The preprocessed data were analyzed with the Exponential Forecasting method, the K-Means statistical analysis, a neural network, and the Recurrence Plots. For estimating future operating conditions, four cases were defined: assuming no significant changes in the vibration amplitudes, a 50% amplitude increment, a 100% amplitude increment, and a 200% increment. The four methods show advantages and disadvantages since they rely only on historical data. It is necessary to gather more data to improve the process for forecasting future conditions since they are easily implemented, but the difference among them requires field experience.