Model-based condition monitoring for wind turbines

摘要: It is common for wind turbines to be installed in remote locations on land or offshore, leading difficulties routine inspection and maintenance. In addition, these are often subject harsh operating conditions. These challenges mean there a requirement high degree of Consequently, monitoring diagnostics play an increasingly important role the competitive operation farms. The data generated by systems can used obtain models under different conditions, hence predict output signals based known inputs. By comparing obtained from operational with those predicted models, it possible detect changes that may due presence faults. This paper discusses model-based condition methods turbines, which relationships between measured variables modelled using linear artificial neural networks identified acquired turbines. also non-linear state dependent `pseudo' transfer functions. Although parameter have been extensively as basis controllers, research described here represents first occasion they employed system. found network-based outperform models; however, computing power required latter considerably less. Finally, develop adaptive threshold rules critical signals, forming early warning