IMPROVING WIND FARM OPERATION PRACTICE THROUGH NUMERICAL MODELLING AND DATA MINING Francesco Castellani University of Perugia - Department of Industrial Engineering castellani@unipg.it Davide Astolfi University of Perugia Department of Industrial Engineering davide.astolfi.green@gma il.com Ludovico Terzi Sorgenia Green ludovico.terzi@sorgenia.it Emanuele Piccioni University of Perugia - Department of Industrial Engineering piccioni@mec.dii.unipg.it Abstract Knowledge of the wind energy production potential is fundamental both for the development and construction of a new wind farm as well as during the operational phase, especially in the early period. Usually wind energy assessment is carried out with the maximum effort prior the installation of the turbines using both numerical and experimental investigations. Anyway, despite the very fast development of the assessment techniques, too many times the actual wind farm performances during operation are not as good as expected according to the previous estimations. This issue can be fixed not only pushing the wind assessment methods to the maximum reliability standards but also with a deep analysis of the operational conditions of the wind farm. The large amount of data collected by the SCADA (Supervisory Control And Data Acquisition) systems installed on the turbines can be very helpful. In present work the performances of a wind farm were analysed through the elaboration of SCADA data from a wind farm in southern Italy; in this site Sorgenia Green installed nine aerogenerators with a rated power of 2 MW on an hilly area with quite gentle slopes. All the most important turbine parameters were used not only for the performance analysis but also to enrich the operational practice trying to define the wind farm behaviour especially in the main direction sectors. This goal was pursued basically with data mining and using wind flow numerical modelling. Keywords: SCADA, condition monitoring, wind farm performance, data mining. 1 Introduction The expertise on wind farm operation is a very important skill that can be fundamental to improve the actual performances in the energy conversion. Many useful information on the expected wind farm response to the wind can be found in the wind energy assessment study through the analysis of measurements and numerical simulations. Anyway a very deep operational analysis needs to be enriched by data mining of all available information spreading out from the turbine’s control systems. In this work a new method of performance analysis using both experimental data as well numerical techniques was tested on a small wind farm operating onshore on an hilly terrain. This kind of approach allowed unveiling all the fundamental information that can be available from operational SCADA data and that could be very useful in order to schedule maintenance intervention in an optimal way or to plan wind turbine stops when power regulation can be useful. 2 The Wind Farm This work deals with the production and wind data of a wind farm located in the south of Italy: it is located in a quite flat land where there are installed nine REpower turbines with a specific rated power of 2000 kW, a hub height of 80 m and a rotor diameter of 82 m. Sorgenia Green, the wind farm owner, provided a database of SCADA and anemometer data that was analyzed and compared with results from numerical simulations. In figure 1 the wind farm layout is represented.