Two-tier model for wind turbine fatigue assessment based on SCADA-dependent neural networks

As older wind farms begin to reach their design lifetime, topics such as wind turbine fatigue, the remaining useful lifetime and lifetime extensions also start to be discussed with greater frequency. It is in this context that wind turbine fatigue assessment appears as crucial contribution. Its accurate portrayal can enable informed decisions regarding lifetime extensions. Moreover, improvements in fatigue assessment have an impact in offshore wind farms’ design, as this is currently fatigue-driven. Fatigue might be derived from strain gauges but, when one considers the prevailing paradigm, their installation on all turbines of a given farm isn’t feasible due to its expensive and time-consuming nature. To circumvent this issue, supervisory control and data acquisition (SCADA) and accelerometer data might be used. This would reveal itself to be an attractive alternative, as the SCADA data in need is already being captured by turbines following the industry’s standard and accelerometer installation is cheaper than strain gauges. In the present contribution, an offshore wind turbine fatigue assessment is carried out through the employment of a two-tier model, consisting of two independent neural networks, wherein the last estimates the tower fore-aft bending moment damage equivalent moment (DEM). The results are finally validated against the DEM of a real-life wind turbine, obtained through the rainflow counting of strain measurements.

Jaar van publicatie:2021

Toegankelijkheid:Closed