ORCID

Abstract

The transformation from fossil fuels to renewable energy is gaining momentum. Countries worldwide are investing significant funds and resources in this shift. The motivations for this change include political, environmental, and social factors, among others. In the UK, which is surrounded by the ocean and experiences strong winds, offshore wind farms are seen as the most viable source of renewable energy. It is anticipated that by 2030, all electricity in the UK will be generated by wind turbines. Hence, the optimisation of the wind farm layouts is the desired direction of research in the UK.In Chapter 3, we discussed the wind farm optimisation problem proposed by Mosetti et al. (1994), which remains a benchmark in the field today. They introduced three scenarios under different wind conditions and used genetic algorithms to solve them. By applying two types of algorithms, deterministic (hill-climber) algorithms and multi-objective evolutionary algorithms (specifically, the Non-dominated Sorting Genetic Algorithm II (NSGA-II), Strength Pareto Evolutionary Algorithm 2 (SPEA2), and Pareto Envelope-based Selection Algorithm II (PESA-II)), we found the optimal solution for the second scenario. We used the hypervolume indicator to assess the performance of evolutionary algorithms (EAs) and plotted their progress as a function of the number of function evaluations using boxplots. This approach provided a better understanding of the optimisation process and should serve as a reference for future publications.In Chapter 4, we solved a six-objective optimisation problem that falls within the category of many-objective optimisation. This category is under-represented in the literature, and several significant gaps exist. Firstly, the visualisation of non-dominated solutions is quite limited. Therefore, in this thesis, we propose to utilise various visualisation techniques, including pairwise plots, parallel coordinate plots, and dimension reduction methods such as Principal Component Analysis (PCA) and Multidimensional Scaling (MDS). To evaluate the performance of evolutionary algorithms (EAs), we em-ployed three different performance indicators: hypervolume, spacing, and generational distance, monitoring their progress in relation to the number of function evaluations.In Chapter 5, we optimised wind farm layouts for floating wind turbines. This is a relatively new topic, as the vast majority of wind farms are fixed and considerably cheaper to operate. However, in the future, when fixed wind farms occupy shallow waters, we may be forced to utilise deeper waters.In the first part of the chapter, we utilised real data from Equinor to model the drift of wind turbines driven by wind, waves, and currents. We found that the linear regression model outperformed both the Random Forest and the Multilayer Perceptron models. However, we concluded that insufficient data are available for a comprehensive experiment.In the second part, we focused on a multi-objective optimisation problem: minimising the perimeter while maximising the wind farm’s power output. Our analysis demon strated that, with respect to these objectives, floating wind farms can be a more advantageous choice than fixed ones. This thesis offers an insightful perspective on the optimisation of wind farm layouts. It begins by introducing the theoretical aspects of optimisation, ranging from single to many-objective optimisation. Additionally, it proposes a variety of visualisation techniques and discusses the topic of floating wind turbines, which could represent a significant advancement over current technologies

Awarding Institution(s)

University of Plymouth

Supervisor

Matthew Craven, David Walker

Document Type

Thesis

Publication Date

2026

Embargo Period

2026-05-01

Deposit Date

May 2026

Creative Commons License

Creative Commons Attribution-NonCommercial 4.0 International License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License

Download

Share

COinS