Improvement of low frequency identification for wind turbines employing eemd and time integration.

Abstract

Nowadays wind turbines are wide employed as a clean energy resource. However, their use implies a wide number of eolic structures which also requires a demanding monitoring and maintenance. The most technique employed for those kind of tasksis vibration analysis inasmuch as it is non-destructive technique and presents a high performance for this task. Among vibration analysis, Components identification turns important because it allows programming the machine maintenance in a proper time. In addition, the accessibility for installing several mechanical vibration transducers into the machine is limited either by the physical space or the high cost of sensor networks. Therefore, it is useful and needed to perform an analysis using just one sensor. In that sense, the accelerometers are commonly utilized since these sensors allow extracting the velocity and displacement, signals which have additional information about the machine, performing a double digital integration. Nonetheless, digital integration evolves several difficulties such as biased errors, leakages in the signal, and strong instability at low and high frequencies.This paper proposes a new methodology based on the ensemble empirical mode decomposition (EEMD) for extracting low frequency components of a wind turbine structure from a single-channel vibration measurement and double integration in time to improve the interpretability of the components.The methodology overcomes the lack of multiple vibration measurements using pseudo-sources and it is especially suitable forsignals with high frequency behavior.