Supervisory Control And Data Acquisition (SCADA) systems have recently become ubiquitous in wind energy technology. SCADA data analysis actually can provide considerable performance improvement at low cost. This also boosts wind energy exploitation, because it enlarges short and long term economic sustainability of investments. Nevertheless, SCADA data analysis poses several scientific and technological challenges, mostly related to the vastness of the data sets required for significant analysis. Separating the signal from the noise is therefore a complex task. In the present work, this issue is tackled by the point of view of state dynamics of wind turbines. SCADA control systems often record superabundant and ambiguous information. Therefore, in this work it is shown that hierarchical classification of information and time discretization of the continuous motion of states are powerful tools. The time-discretized state dynamics is processed in the formulation of several indices for performance evaluation and fault diagnosis. The method is tested on the data set of a wind farm owned by Renvico s.r.l. and sited in Italy.