| Title | The Overlooked Foundation: The crucial Role of Preprocessing in Enhancing rock units characterization |
|---|---|
| Authors | Hammoud A, Villeneuve M, Caspari E, Mortensen A, El Kasti K |
| Year | 2026 |
| Conference | World Geothermal Congress |
| Keywords | |
| Abstract | As data science continues to gain ground in geothermal exploration and drilling, there is growing emphasis on automation and machine learning. However, the foundational step of data preprocessing remains underappreciated despite being the most time-consuming and critical phase. Before any meaningful interpretation or model output can be achieved, raw data must undergo a rigorous transformation process involving selection, resampling, scaling, and careful evaluation of anomalies and heterogeneities. These preprocessing stages are not just preliminary—they are essential for ensuring the quality and integrity of any subsequent analysis or automated workflow. This paper explores the role of data preprocessing using case studies from wireline logging and drilling data collected while drilling production wells at two geothermal fields in NE-Iceland. A key aspect of data selection is setting a clear objective, which guides the selection of relevant parameters and the identification of what could affect these parameters. Given that datasets are often recorded at varying resolutions, resampling to a unified basis—typically depth—is crucial. Additionally, many machine learning algorithms are scale-sensitive, necessitating for example proper normalization. Importantly, not all outliers are errors; some anomalies reflect meaningful geological events and should be retained. In geothermal exploration, success begins with how we prepare our data—because precision at the start determines performance at the end for better geological understanding and smarter decision making. |