A Transition from Deterministic to Stochastic Cost Models for Offshore Wind Farms -The End of "Best, Worse and Average" Scenarios

Esteve Borrās Mora 1 ,2, James Spelling2, Harry Van der Weijde3
1Industrial Doctoral Centre for Offshore Renewable Energy (IDCORE), Edinburgh, UK, 2EDF Energy R&D UK Centre, London, UK, 3University of Edinburgh, Edinburgh, UK


Offshore wind cost modelling seeks to understand and quantify how different project specifications, technology choices and market trends contribute to the overall project finances. Better understanding of costs leads to more informed business decisions in the industry that will help shorten the gap between engineering and financing. In order to carry out such tasks, project developers, investors and bankers need to get a better understanding of the risks that they are faced with: Will the wind blow as predicted? Will the export cable have the stated availability? Will construction and commissioning be completed on time without any major overrun? Will the O&M strategy be performed as described?
Often, scenario analysis has been used in order to address complicated cost models for offshore wind farms, by means of best, worse and average scenarios. Differences between them were used as a measure of risk by decision makers. However, in general, such analyses are not very informative and differ substantially from the risk measures used in the financial world. Establishing a probabilistic framework helps to better quantify these risks, identify how they impact the costs and allow engineers and investors to "talk the same language" about common risk measures.


To answer all these questions, EDF R&D's Offshore Wind Cost Analysis Tool (OWCAT) is able to rapidly evaluate the financial performance of a project, based on the selected project parameters and local market conditions, helping to quickly compare different types of wind turbines, foundations, electrical configurations, etc. and how these technical specifications are translated into financial metrics, all of them with their associated uncertainties. The methodology that has been proposed is illustrated using a series of standard test cases, focusing on different levels of uncertainties and technologies choices.


Two models were used and compared against each other. On the one hand, a deterministic model was used to estimate the financial metrics, assuming fixed values as inputs (the expected values). On the other hand, the stochastic version of the model was utilized to obtain a new set of financial metrics within a probabilistic framework. The input probability distributions and interdependencies have been obtained from discussions with internal experts as well as publicly available data, providing a framework for stochastic modelling. This has proved very valuable, as it effectively summarizes and brings together various, distinct consequences associated with different phases of the wind farm project.


This innovative way of quantifying uncertainty involved all experts feeding sound estimates into an integrated cost model. A sound estimate is considered to be such that it provides information on the probabilistic distribution. Topics covered range from project development to project financing, going through wind resource assessment, wind turbine technology, geotechnical and foundations and electrical system, among others. By embedding Uncertainty Quantification (UQ) into OWCAT, resulting into quantitative answers to the financial and technical community, a substantial improvement in the performance of the tool has been reached in comparison to the deterministic approach. The paper concludes with and highlights the importance of undertaking stochastic cost modelling. In a nutshell, it tells us that dealing with uncertainty really matters.


During the course of this presentation EDF Energy R&D UK Centre will cover the following topics:
Methodology on how to transition to the stochastic models for added value results and deeper insights on cost modelling to provide developers with an advance decision-making tool
Analysis of different generic test case studies