Wind Speed Prediction Feature Design for Echo State Networks

de Jong, Niels (2020) Wind Speed Prediction Feature Design for Echo State Networks. Bachelor's Thesis, Artificial Intelligence.

Preview	Text AI_BA_2020_Niels_de_Jong.pdf Download (1MB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (94kB)

Abstract

Due to a multiplicity of factors, the world's energy demand will increase for the upcoming decades. In meeting this demand, renewable energy sources such as wind are seen as attractive alternatives, but their high degree of variability makes integration into power grids a challenge. Intelligent prediction methods, such as echo state networks (ESNs), may prove to be part of the solution. However, studies on feature design for input into such networks is scarce. Thus, the question "What good features can be designed in helping to solve the very short-term wind speed prediction task by supervised machine learning?" is posed, concentrating on ESNs. To address this question, six meteorological variables from the WIND toolkit are subjected to a series of performance tests via appropriately tuned ESN models. By comparing resulting average NRMSE values to a reference persistence model, three conclusions are drawn: (I) only the feature wind speed appears to be informative in the task; (II) spatially diverse features are informative, and (III) decomposing input features into smooth and noise sub-signals raises informative quality.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Jaeger, H.
Degree programme:	Artificial Intelligence
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	02 Jul 2020 13:40
Last Modified:	02 Jul 2020 13:40
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/22341

Actions (login required)

View Item