Geothermal energyProject Data Fusion and Machine Learning
A report on an algorithm for quantifying the inherent uncertainty in the geothermal exploration problem enabling geothermal investors and developers reason about the risk of taking a certain action or the value of financing a particular survey.
The inference engine is an algorithm for performing multi-sensor, Bayesian, geophysical inversions using analytic and numerical forward models and strong geological prior information. It produces posterior probability distributions over 3D geological structures that can be queried for critical geothermal variables such as rock type and temperature. In principle, it is also capable of inferring and representing fracture levels and other physical rock properties.
We have formulated an approach that provides a probabilistic framework for quantifying the inherent uncertainty in the geothermal exploration problem. Bayes’ rule enables the algorithm to update our prior beliefs about the geology, based on the information from a multitude of sensors and sensor types. The output of the algorithm is a full probability distribution over possible geological models rather than a single solution. This distribution enables us to reason about the risk of taking a certain action or the value of financing a particular survey.