Back to search

University of Glasgow

Development and applications of AI for offshore wind power forecasting

Abstract

dc:description.abstract

Wind energy plays a vital role in securing a sustainable and low-carbon future, strengthening energy independence, enhancing economic growth, and preserving the environment. In addition to reducing climate change impacts, wind power is able to facilitate the development of a more resilient and sustainable energy system. There is one obstacle, though, that prevents its penetration into the power grid: its high variability in terms of wind speed fluctuations. Wind power forecasting plays a vital role in addressing the inherent uncertainty of wind power generation. Accurate power forecasting, while making maintenance more efficient, leads to profit maximisation of power traders, whether for a wind turbine or a wind farm. Several studies have been conducted in the past to investigate factors affecting the performance of power forecasting methods, and several models have also been developed. It is, however, necessary to develop a method that not only provides high prediction accuracy, but also provides good efficiency as well. This thesis explores different forecasting approaches for wind energy and uses machine learning to develop an accurate, efficient, and robust prediction model. First, background and literature review is presented which covers analysis methods, forecasting time scales, error measurement, and accuracy improvement. Following this, in order to provide high-quality and noise-free data for wind power forecasting, several preprocessing techniques were investigated. Next, the research focused on fine-tuning the hyperparameters of machine learning models to increase forecasting accuracy and efficiency. Scikit-opt, Hyperopt, and Optuna, three hyperparameter optimisation techniques, are used to tune CNNs and LSTMs, two commonly used deep learning models. After analyzing the results of the previous sections, a new wind power forecasting method is proposed using Wavelet Packet Decomposition (WPD) models, optimised LSTM models and CNN models. After preprocessing the raw data and removing the outliers, WPD is employed to decompose wind power time series into multiple subseries with different frequencies. Comparing the prediction results of all involved models demonstrates that the developed model improves the prediction accuracy by at least 77.4% compared to methods that do not use WPD. In addition, the proposed combination of optimised CNN and LSTM improves the forecasting accuracy by 26.25% compared to methods that use only one deep learning model to forecast all sub-series. In light of the success of the one step ahead forecast, different strategies of multi-step ahead forecasting were explored for the first time in the field of wind power forecasting. The results show that in twostep ahead wind power forecasting, all strategies produce similar results, in both wind turbines. In all forecast horizons of more than two steps ahead, the MIMO approach is best when the dataset does not contain any outliers. In contrast, the direct approach is best when the dataset does contain outliers. It was also concluded that when datasets contain outliers, wind power forecasting using recursive strategies results in the highest errors for forecasts over two steps.

Degree

thesis:*
Level dc:type.qualificationlevel
PhD
Grantor dc:publisher.institution
University of Glasgow
Year dc:date.issued
2024

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Hanifi, Shahram

Subjects

dc:subject × 1

Rights

Language dc:language
en

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
University of Glasgow
Base URL
theses.gla.ac.uk/cgi/oai2
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Hanifi, Shahram. Development and applications of AI for offshore wind power forecasting. PhD thesis, University of Glasgow, 2024. https://doi.org/10.5525/gla.thesis.84539