Deep Learning for Geothermal Reservoir Characterization: Estimating Rock Properties from Seismic Data Using Convolutional Neural Networks

Abstract

Abstract Estimating rock properties is a crucial aspect of geothermal reservoir characterization, which plays a pivotal role in the efficient harnessing of geothermal energy. Rock properties include hydraulic properties, such as porosity and permeability, and elastic properties such as Poisson’s ratio, P-wave, S-wave velocity, bulk modulus, and acoustic impedance. Accurate determination of these properties allows geoscientists and reservoir engineers to assess and optimize the reservoir performance and assess the long-term stability of geothermal projects. Seismic inversion is the process of deriving these rock properties from seismic data. Conventional seismic inversion can be time-consuming and costly. Machine learning can effectively estimate rock properties which reducesthe need to rely on conventional seismic inversion, expensive lab experiments, and well logging data. This study aims to estimate keyrock properties (acoustic impedance, bulk modulus, density, permeability, Poisson’s ratio, and porosity) from the SCAN dataset using a convolutional neural network. The proposed U-net architecturewas used to develop models that rely on a full-stack seismic dataset as inputs to the model. Mean SquaredError (MSE) with a regularization factor was considered as a loss function when training the model and Mean Absolute Error (MAE) to assess the performance of the model. Results reveal an effective performance of the developed models in the estimation of rock properties with low MAE values ranging between 0.5-3 %. The higher MAE observed for the porosity and permeability estimation is attributed to poor data coverage in the ground truth data.This study demonstrates the potential of convolutional neural networks to predict rock properties from seismic data for efficient reservoir characterization.