Model-Based Load Estimation for Predictive Condition Monitoring of Wind Turbines

The main objective of this paper is to present a Load Observer Tool (LOT) for condition monitoring of structural extreme and fatigue loads on the main wind turbine (WTG) components. LOT uses well-known methods from system identification, state estimation and fatigue analysis in a novel approach for application in condition monitoring.

Fatigue loads are estimated online using a load observer and grey box models which include relevant WTG dynamics. Identification of model parameters and calibration of observer are performed offline using measurements from WTG prototype. Signal processing of estimated load signal is performed online, and a Load Indicator Signal (LIS) is formulated as a ratio between current estimated accumulated fatigue loads and its expected value based only on a priori knowledge (WTG dynamics and wind climate). LOT initialisation is based on a priori knowledge and can be obtained using a high-fidelity aero-elastic simulation code as LACflex to increase LIS rate of convergence. Additionally, statistical uncertainties in the convergence of estimated fatigue loads due to climate parameters are taken into account in the LIS function. LIS may play a central role in condition maintenance programme for pre-maintenance actions.

The performance of LOT is demonstrated by applying it to one of the most critical WTG components, the gearbox. Model-based load CMS for gearbox requires only standard WTG SCADA data. Direct measuring of gearbox fatigue loads requires high cost and low reliability measurement equipment. Thus, LOT can significantly reduce the price of load monitoring.