Machine Learning in GeoEnergy Systems is a course bringing machine learning into GeoEnergy engineering, which leverages different machine learning techniques for solving GeoEnergy engineering problems.

There exists the high dimensionality of data in GeoEnergy engineering coming from different geoscience and engineering disciplines, for instance, 1D well-logging data, 3D geo-cellular-based properties (permeability, porosity, Youngs Modulus), time series of production or injection data, and temporal-spatial (3D or 4D) data of pressure, saturation, temperature, and stress fields. The prediction of these data is usually based on physics laws, such as fluid flow in porous media, wellbore hydraulics, etc., complicating the prediction tasks with low efficiency and significantly decreasing the decision-making processes. In this course, we will illustrate how to efficiently perform exploratory data analysis and visualization of these GeoEnergy data, find correlations between different features for targeted prediction tasks based on the inherent physics of flow in porous media, perform dimensionality towards high-dimensional input data (e.g., geo-cellular data) and high-dimensional output (e.g., temporal-spatial reservoir dynamic data), and generate the most important features through sensitivity analysis. Further, we will introduce different machine learning and deep learning models (supervised or un-supervised) to perform the predictions (e.g., reservoir simulation with deep learning), and compare the prediction results with experiments, field observation or monitoring data, and data generated by reservoir numerical simulation. Moreover, some optimization techniques like gradient-based or gradient-free optimization will be taught to couple with these machine learning models to deliver the optimum reservoir control parameters while maximizing the GeoEnergy reservoir performance. Finally, special topics seminars for machine learning applications in predicting geothermal recovery, geological carbon sequestration, etc, will be introduced to help students catch the latest advances in our domain.