Détection de sous-performances de centrales éoliennes avec l’aide de cartes de contrôle

Abstract

The main objective of this thesis is to develop methods to monitor wind farms. This objective fulfills the actuals needs of wind farm operator. With the ageing of their assets, wind farms operators are concerned with the condition of their wind turbines. The developed methods aim to be applicable in an industrial context. Thus, they are based on operational data. There is no need to add more sensors on the wind turbines or to do regular inspections, hence facilitating the deployment of the proposed methods. The use of operational data increases the scope of the work achieved in this thesis by making it applicable to almost all the wind farms in operation. Three wind farm monitoring methods have been developed. The first one follows the wind turbines performance with the monitoring of the power curve. First, the power curve is obtained with an approach inspired by the bins method defined in the IEC 64100-12-1 standard. Then, the residuals between the observed and the expected power are analyzed with an Exponentially Weighted Moving Average (EWMA) and a Generally Weighted Moving Average (GWMA) control chart. The second method is for the monitoring of the wind turbine main bearing. This method uses a physic model based on heat transfer. A badly conditioned bearing will emit more heat due to friction augmentation. As in the first monitoring method, an EWMA control chart is used to detect such issues. Two case studies are presented as for an illustration of the industrial potential of this monitoring method. The last monitoring method developed is a data-driven approach. This method uses the difference between a single wind turbine and the average of the other wind turbines of the wind farm for each quantity measured on the wind turbines (power, azimuth, component temperature, etc.) An EWMA control chart is used to establish if the wind turbine suffered any behavior changes. This thesis arises from an industrial collaboration. The data used for the development and the validation of the monitoring methods are from Cartier Énergie Éolienne’s wind farms. Through the development and the implementation of the monitoring methods, a certain number of issues regarding the wind-turbine performance or the condition of the wind turbines and their components have been reported. The identification of issues assesses the industrial potential of the various methode developed achieved in this thesis.

Date	13 Jun 2017
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Christian Masson (Supervisor) & Antoine Tahan (Co-supervisor)