Computational science

Research

How much wind power models can be trusted, and what changes downstream when their inputs are corrected. My PhD corrected ERA5 reanalysis wind against observed generation, then carried the correction into European energy system models to measure how wake losses, bias and spatial resolution move investment decisions.

01

Thesis

Oct 2021 – Jun 2026EPSRC and Shell Industrial CASE award

Data Science-Enhanced Wind Power Modelling: From Reanalysis Correction to Energy System Representation

PhD in Computational Science & Data Science, Imperial College London

  • Gridded corrections from 1,729 control points in 12 countries reduced capacity-factor MAE by 16 to 79% against uncorrected ERA5.
  • Validated against ENTSO-E generation for 2023, the corrected model improved 9 of 11 country and technology series, cutting NRMSE by up to 36%.
  • Wake losses reduced optimal offshore wind capacity by 36 to 46% at every scale studied.
  • In a seven-country North Sea system, spatially resolved correction changed total system cost by 8.1%, more than twice the effect of uniform scaling.
  • Terrain-informed machine learning lost to geometric interpolation under spatial cross-validation. I reported the lower figures and kept interpolation as the published method.

Wind resource representation belongs with the major techno-economic assumptions in energy system modelling, alongside discount rates and cost projections.

02

Publications

Energy · 2024First author

Improving wind power modelling through granular spatial and temporal bias correction of reanalysis data

Energy, 313, 133759

Introduces the granular spatial and temporal bias-correction method that PyVWF implements, trained on Danish turbines and validated on a held-out year with 1,445 new turbines.

View publication
Energy Conv. & Mgmt · 2026Co-author

Geographic variability in reanalysis wind speed biases: a high-resolution bias correction approach for UK wind energy

Energy Conversion and Management, 352, 121066

Applies high-resolution reanalysis bias correction to UK wind energy.

View publication
Applied Sciences · 2022Co-author

Multi-Output Regression with Generative Adversarial Networks (MOR-GANs)

Applied Sciences, 12(18), 9209

Grew out of my MSc dissertation on Wasserstein GANs for multi-output regression, benchmarked against Gaussian process regression.

View publication