Data Driven Fatigue Load Monitoring in a Wind Farm

PhD by Luis Vera-Tudela, Peru (PPRE2015-2017)
Institution: ForWind – University of Oldenburg



Wind turbine lifetime estimation has become an area of interest to reduce the cost of energy and to improve O&M. There are more than 268 thousand wind turbines installed in the world (GWEC, 2014), enough reason to research on how to improve machinery use. Lifetime estimation via the calculation of fatigue loading is one possible approach, which is also followed during the design of wind turbines. However, its continuous monitoring is rather impracticable since it needs extra measurements. In this context, monitoring it using 10-minute statistics of operational (SCADA) signals became an area of research few years ago.

This PhD is based on the previous claim: it is possible to map the relationship between statistics of SCADA signals and fatigue loads with a neural network model, but this is limited to single wind turbines under free stream flow conditions. Goals of the PhD were then to explore, improve and extend the approach for its use in wind turbines located in a wind farm. Measurements from offshore wind farms EnBW Baltic 1 and Alpha Ventus were available for the research.

Results of the PhD (due at the end of 2016) focused on data-mining approach and included: a method to select an optimum -minimum- set of input variables for prediction, which was reduced to a tenth of previous levels; a framework to determine prediction quality, which allows the quantification of future improvements; evaluation of predictions for different wind flow conditions and its assessment in realistic conditions, i.e. on other equivalent turbines in a wind farm. Finally, alternative approaches, like physical, stochastic and hybrid models were investigated, test and reported.

Webmhbdasm73tter440hx ( (Changed: 2020-01-23)