Robust estimation and diagnosis of wind turbine pitch misalignments at a wind farm level

Wind turbine pitch misalignments provoke aerodynamic asymmetries which cause severe damage to the turbine. Hence, it is of interest to develop fault tolerant strategies to cope with pitch misalignments. Fault tolerant strategies require the information regarding the diagnosis and the estimation of the faults. However, most existing works focus only on open-loop misalignment diagnosis and do not provide robust fault estimates. In this work, we present a novel strategy to both estimate and diagnose pitch misalignments. The proposed strategy is developed at a wind farm level and it exploits altogether the information provided by the temporal and spatial relations of the turbines in the farm. Fault estimation is first addressed with a closed-loop switched observer. This observer is robust against disturbances and it adapts to the varying conditions along the wind turbine operation range. Fault diagnosis is then achieved via statistical-based decision mechanisms with adaptive thresholds. Both the observer and the decision mechanisms are designed to guarantee the desired performance. Introducing some restrictions over the number of simultaneous faulty turbines in the farm, the proposed approach is ameliorated via a bank of the aforementioned observers and decision mechanisms. Finally, the strategies are tested using a well-known wind farm benchmark.