Analyzing state dynamics of wind turbines through SCADA data mining

Supervisory Control And Data Acquisition (SCADA) control systems have become ubiquitous in modern wind energy technology. Exploiting their potentialities is a keystone for performance optimization and to improve the operational control feeding smart electric grids. Yet, tackling the complexity of SCADA data sets is a challenging task. The philosophy underlying the present work is discretization of the continuous motion of machine states and error signals: Doing this, one ends up with simplified databases, acting on which with statistical methods provide powerful insight. Indicators on the quality of turbine functionality and on the nature of error signals are formulated, and the distribution of errors as a function of wind intensity is studied. The methods are tested on the data of a wind farm in southern Italy, and it is shown that they are indeed capable of assessing performances and interpreting the nature of occurred errors.