Mathematical Methods for Performance Evaluation of Onshore Wind Farms through SCADA Data Mining Francesco Castellani University of Perugia Department of Engineering Italy francesco.castellani@unipg.it Davide Astolfi University of Perugia Department of Engineering Italy davide.astolfi.green@gmail.com Ludovico Terzi Sorgenia Green srl Via Viviani 12, Milano Italy ludovico.terzi@sorgenia.it Abstract: Conversion of wind energy into electricity is affected by low density of wind turbines on the territory and by the short term stochastic nature of the source. In order to build smart electric grids, and therefore to balance vari- able loads efficiently with not renewable sources, sophisticated control systems and optimization techniques are the keystone. In the present work, data spreading out from wind turbines Supervisory Control And Data Acquisition (SCADA) control systems are exploited for the issue of performance evaluation and to build a connection between performances and mechanical behaviour of the machines. The philosophy of our post processing is zooming into the regime of unison power output production of the wind farm and simplifying appropriate amounts of informa- tion into a binary map, from which extracting explanatory answers is easier and more powerful. Our methods are tested on onshore wind farms owned by Sorgenia Green in southern Italy, and it is shown that they are capable of performance evaluation and insight into mechanical aspects of wake interactions between nearby turbines, as rotor misalignment with respect to wind direction and consequent production loss. Key–Words: Wind energy, wind turbines, SCADA control system, performance evaluation. 1 Introduction Optimization of operating wind farms has become a fertile field in the latest years, due to the financial crisis which has led to a slowdown of new installa- tions and because a remarkable amount of energy dis- patched into the electric grids originates from stochas- tic sources, as wind energy is. Therefore, in order for the grid to be as “smart” as possible and to balance efficiently with energy com- ing from not renewable sources, sophisticated control systems are needed, especially for energy conversion systems, as wind turbines, having low density on the territory. Exploiting control systems potentialities is useful both for optimization and for fault prevention: in [1] it is estimated that a sudden failure of a 1.5 MW wind turbine during winter time leads to around e50000 of missed production. This amount is up to 5 times greater than the missed production due to a wisely planned maintenance program. Similarly, in [2] it is estimated that a complete optimization program of wind turbines might diminish up to a 20% the cost of energy generated from wind. Sophisticated Supervisory Control And Data Ac- quisition (SCADA) have rapidly become a major sub- ject in the scientific literature, not only for the eco- nomic advantages above depicted, but also, and most valuably, because codifying and extracting informa- tion on the machines from the output of control sys- tems is a challenging task. SCADA control systems spread on 10 minute time basis minimum, maximum, average and standard deviation of a vast amount of measurement channels, regarding the details of the wind flow, machine re- sponse and alignment to it, the conversion of wind kinetic energy into active power, the vibrational sta- tus, the temperatures at the most peculiar parts of the turbine. Vast data streams must be turned into informa- tion, further concentrated and processed into knowl- edge, and possibly integrated in the control system itself for automation of early diagnosis and perfor- mance evaluation techniques. Methods at this aim basically demarcate in two categories: statistical approaches on post-processed data sets and numerical modeling, especially Artificial Neural Networks techniques because of their capabil- ity in codifying non-linearity. For a comprehensive review of the different approaches, based on statistics, physics, modeling and data mining, we refer to [3]. In [4] an evolutionary strategy algorithm is used for solving optimization models and to determine op- timal control settings: it is shown that judicious fine tuning of blade pitch angle and generator torque im- Recent Advances in Energy, Environment and Financial Planning ISBN: 978-960-474-400-8 105