Enhanced Kalman Filter Through Modified Empirical Mode Decomposition For Wind Profile Exogenous Disturbance Extraction & Isolation in Wind Turbines

Abstract

Wind profile variations and disturbances are the main cause for stress and fatigue for wind turbines. These disturbances propagate along the drive train, through the gearbox and into the generator resulting in current and voltage output fluctuations. The wind profile is a non-stationary random process, thus the resulting vibrations and disturbances throughout the system are non-stationary. Classical traditional frequency-domain analysis techniques fall short when dealing with this type of signals. Modern analysis and control requirements in wind turbines justify the need for advanced techniques to cope with the non-stationary nature of measured signals. Compensating these disturbances to protect different wind turbine components, while detecting harmonics caused by these disturbances, render the turbine system operation smoother while increasing reliability, efficiency and robustness. This paper applies a Kalman filter based method for signal reconstruction through harmonic estimation for the turbine side angular velocity. In addition, a new modified Empirical Mode Decomposition (EMD) approach is introduced capable of separating the continuous component of a non-stationary signal from its added high and low frequency waves. The modified EMD intends to reduce time consumption for signal processing and isolate harmonics from the carrier wave in the angular velocity signal for analysis. Then the EMD and the Kalman filter are combined in order to improve individual harmonic component estimation while allowing the use of conventional signal processing techniques. The method can be used either to reject wind profile disturbances, or detect added fault signatures by a single component.