| Title | Power Density Geothermal Resource Estimation Revisited |
|---|---|
| Authors | R. Chadwick HOLMES |
| Year | 2024 |
| Conference | Stanford Geothermal Workshop |
| Keywords | resource estimation, power density, power per drill length, analogs, data analytics, machine learning |
| Abstract | The power density method simplifies geothermal resource assessments to just reservoir temperature, resource area, and tectonic environment, using trends observed in a collection of global analog fields to predict power potential in undeveloped areas. For geothermal explorers, the procedural simplicity of power density over other resource estimation methods holds great attraction. However, the degree of variability in the field data used to define power density relationships suggests a non-trivial level of uncertainty in predictions made using traditional power density curves or related resource calculators. Power density also lacks a clear linkage to factors like drilling costs that greatly influence geothermal project economics. This study revisits the power density method by first evaluating its predictive performance in its three-variable form and then expanding the field data using multiple global databases. Representative values for subsurface characteristics like reservoir temperature and target depth are combined with features describing the surface plant design and climate conditions to paint a more complete picture of each field location. The data are analyzed to reveal the most important features for predicting power production in aggregate by field and on an individual power plant basis. New models created from these features maintain a low level of complexity appropriate for exploration. Furthermore, they are easily tuned to predict novel power metrics like power per drill length that incorporate project economic drivers to aid in geothermal strategy and portfolio-building efforts. |