Bayesian neural networks for geothermal resource assessment: Prediction with uncertainty

Abstract

In this work, the application of machine learning to the evaluation of geothermal resource potential is considered. A supervised learning problem is defined where maps of ten geological and geophysical features within the state of Nevada, USA, are used to define geothermal potential across a broad region. With the availability of a relatively small set of positive training sites (known resources or active power plants) and negative training sites (known drill sites with unsuitable geothermal conditions) two classes of artificial neural networks, for this classification task, are constrained and optimised. The main objective is to predict the geothermal resource potential at unknown sites within a large geographic area where the defining features are known. These predictions could be used to target promising areas for further detailed investigations. The work evolves from defining a specific neural network architecture to training and optimisation trials. The inevitable problems of model variability and resulting prediction uncertainty are exposed. Finally, to address these problems, the concept of Bayesian neural networks, a heuristic approach to regularisation in network training, is exploited, together with the practical interpretation of the formal uncertainty measures they provide.