Wave-by-wave Forecasting of Sea Surface Elevation for WEC Applications Utilizing NARX Neural Networks Public Deposited

http://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/zw12z866q

Descriptions

Attribute NameValues
Creator
Abstract or Summary
  • Forecasting of ocean waves over a short duration on the order of tens of seconds was approached with the optimization of wave energy conversion in mind. This study outlines the development of an artificial neural network model, specifically the Nonlinear Autoregressive Network with Exogenous Input (NARX), to predict a wave-by-wave surface elevation time series based entirely on previous observations at the site of interest. Such a model would be computationally less intensive than competing deterministic techniques rooted in wave theory. Furthermore, it could potentially fit irregular patterns without some predetermined function as would be necessary for many other stochastic approaches. In principle, neural networks can be trained to learn previous patterns based on a recent wave record and then utilized in a feedback mode to yield multistep predictions for perhaps up to three wave periods (~45 seconds). The challenge of this approach is error accumulation in the intermediary steps that can lead to poor performance for longer prediction horizons. It was hypothesized that filtering the wave record input via wavelet or Fourier transformations could enhance model performance and hence, explorations of these types of input preprocessing were an integral part of the study. The NARX architecture allowed for an exogenous input time series that could conceivably mitigate error accumulation by providing an additional degree of signal correlation. Accordingly, the investigation also included potential exogenous inputs that could be derived from the original wave record. The work culminated with a band-pass exogenous input delivering a significant forecasting advantage. Yet, providing the zero-phase narrow-band signal posed a challenge when applied in real- time, without the use of future data. A two-prong tactical approach was undertaken to address this challenge but ultimately proved insufficient. Consequently, the success of the NARX wave-by-wave forecasting method under the conditions of a real-world application will depend upon a better solution to the zero-phase filtering challenge.
Resource Type
Date Available
Date Copyright
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Non-Academic Affiliation
Keyword
Subject
Rights Statement
Peer Reviewed
Language
Replaces
Additional Information
  • description.provenance : Approved for entry into archive by Laura Wilson(laura.wilson@oregonstate.edu) on 2015-12-24T00:17:37Z (GMT) No. of bitstreams: 1 GillespieAliceG2015.pdf: 10454771 bytes, checksum: 45ac3813d001b4a85707d23d58e20f11 (MD5)
  • description.provenance : Made available in DSpace on 2015-12-24T00:17:37Z (GMT). No. of bitstreams: 1 GillespieAliceG2015.pdf: 10454771 bytes, checksum: 45ac3813d001b4a85707d23d58e20f11 (MD5) Previous issue date: 2015-12-17
  • description.provenance : Submitted by Alice Gillespie (gillesal@oregonstate.edu) on 2015-12-18T20:34:52Z No. of bitstreams: 1 GillespieAliceG2015.pdf: 10454771 bytes, checksum: 45ac3813d001b4a85707d23d58e20f11 (MD5)
  • description.provenance : Approved for entry into archive by Julie Kurtz(julie.kurtz@oregonstate.edu) on 2015-12-22T19:43:16Z (GMT) No. of bitstreams: 1 GillespieAliceG2015.pdf: 10454771 bytes, checksum: 45ac3813d001b4a85707d23d58e20f11 (MD5)

Relationships

In Administrative Set:
Last modified: 08/17/2017

Downloadable Content

Download PDF
Citations:

EndNote | Zotero | Mendeley

Items