Modelling the stochastic evolution of wind profiles using neural networks
MetadataShow full item record
The aim of this project is to profile a multivariate set of wind speeds corresponding to different heights and composing a vertical speed profile in time. Most importantly, these analysis will be done over a data-set with a 1Hz sampling frequency, which is uncommon for the industry’s standards, using typically 10 or 15 minute averages. The purpose of for such small sampling frequency is to augment the knowledge over current models under different circumstances. To that end, we will use measurements collected at off-shore platforms at the North Sea FINO1 platform, made available to OsloMet for research purposes. The project is divided into two parts. First, we conduct an analysis of the time and spatial patterns of the set of wind speeds using basic statistical and numerical tools. Second, the implementationtation of Deep Learning architectures using the findings from the analysis conducted on part one. These architectures will then be trained and validated with the data at hand. The final goal is to provide a comparative analysis of how these AI-based models capture the fundamental statistical and dynamical features of wind speed profiles and retrieves predictions of future profile states. Such insights could be of interest for better predicting several aspects related to the design, operation, and maintenance of the next generation of wind turbines. Particularly in the estimation of fatigue loads, which is a major aspect in wind farm monitoring and maintenance protocols.