Addressing Data Scarcity in Offshore Wind CAPEX Forecasting: Regression-Based Approach

Reliable offshore wind capital expenditure (CAPEX) forecasting is constrained by limited and unevenly disclosed project data, particularly the small number of projects with both forecasted and actual costs. This study develops and validates a regression-based approach for expanding the available cost dataset under these conditions. A modelling sample of 59 commissioned fixed-bottom European offshore wind projects was assembled, comprising 48 projects with verified forecast–actual cost pairs and 11 additional projects with verified realised CAPEX. A log–log cost model was specified using project capacity, water depth, distance to shore, and a quadratic term for commissioning-year to account for non-linear changes in costs over time. Seven estimation methods were evaluated using leave-one-out cross-validation. Ordinary least squares (OLS) produced the lowest prediction error, with a mean absolute percentage error of 19.9%. The selected model was applied to 38 additional projects with reported forecasted cost but no reported actual cost, generating model-estimated actual cost and corresponding implied cost overruns. This increased the total CAPEX dataset from 59 to 97 projects and expanded the cost-overrun reference-class dataset from 48 to 86 projects. The additional projects produced a mean implied cost overrun of 17.9% broadly consistent with the verified sample and previous offshore wind evidence. The findings demonstrate that validated regression-based estimation can strengthen offshore wind cost datasets under incomplete disclosure and improve the evidence base for Reference Class Forecasting and the development of more reliable offshore wind cost models.