Options
Multivariate simulation of offshore weather time series: a comparison between markov chain, autoregressive, and long short-term memory models
Citation Link: https://doi.org/10.15480/882.4401
Publikationstyp
Journal Article
Publikationsdatum
2022-06-16
Sprache
English
Institut
Enthalten in
Volume
2
Issue
2
Start Page
394
End Page
414
Citation
Wind 2 (2): 394-414 (2022-06-16)
Publisher DOI
Publisher
MDPI
Peer Reviewed
true
In the estimation of future investments in the offshore wind industry, the operation and maintenance (O&M) phase plays an important role. In the simulation of the O&M figures, the weather conditions should contain information about the waves’ main characteristics and the wind speed. As these parameters are correlated, they were simulated by using a multivariate approach, and thus by generating vectors of measurements. Four different stochastic weather time series generators were investigated: Markov chains (MC) of first and second order, vector autoregressive (VAR) models, and long short-term memory (LSTM) neural networks. The models were trained on a 40-year data set with 1 h resolution. Thereafter, the models simulated 25-year time series, which were analysed based on several time series metrics and criteria. The MC (especially the one of second order) and the VAR model were shown to be the ones capturing the characteristics of the original time series the best. The novelty of this paper lies in the application of LSTM models and multivariate higher-order MCs to generate offshore weather time series, and to compare their simulations to the ones of VAR models. Final recommendations for improving these models are provided as conclusion of this paper.
Schlagworte
MLE@TUHH
DDC Class
600: Technik
620: Ingenieurwissenschaften
More Funding Information
This research was supported by the Bundesministerium für Wirtschaft Und Klimaschutz (BMWK.IIC6) under the grant agreement no. 01186400/1, financial code 03EE3001B, for the ProBeKo project (subproject of the SOBeKo project). Furthermore, the second and third authors received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement no. 815083 (COREWIND).
Publication version
publishedVersion
Loading...
Name
wind-02-00021-v2.pdf
Size
1.81 MB
Format
Adobe PDF