Development of an experimental platform dedicated to wind turbine default detection

Abstract

This work is divided into two distinct topics: the characterization of the sensors developed by the microSys laboratory and the analysis of a Horizontal Axis Wind Turbine. The first part was performed on a shaker, where the operating limits of the sensors have been studied. This research has shown unexpected results, since a decrease in the amplitude measured by the sensors was observed very early on the frequency range. The second part of the project was divided into three main parts: a theoretical study of the wind turbine with the creation of a structural and an aerodynamic model of the machine, a modal analysis performed on a single blade and on the entire structure and finally a series of tests carried on in a wind- tunnel. The machine modelling aims to compare the results predicted by the theory with the experimental measures, but also to understand how the turbine behaves when confronted to faulty situations. Using the modal analysis, it was possible to predict the deformation of the blade and the turbine with the frequency and to understand the vibration phenomena observed experimentally. The wind-tunnel test campaign revealed that the first natural frequency of the blade was excited at high rotation speed and high velocity, leading to an important level of vibration. In order to perform the wind-tunnel tests, a software was developed to receive simultaneously the data coming from the sensor, but also the voltage and the current delivered by the machine. The main goal of this project was to focus on the wind turbine defaults and on their detection. The different tools developed preliminary allowed to perform a complete analysis of the structure and to highlight the characteristics proper to each default. Four main phenomena were studied: the misalignment of the wind turbine with the wind, the load loss, the turbulence and the unbalance. Finally, an analysis aiming to highlight each malfunction was performed. Due to the strong acceleration noticed in the rotation of the blades, the detection of the load loss can be easily carried on, but the three other defaults are strongly dependent on the initial conditions, making the detection more complex.