The primary objective of my study is guided by finding alternate low-carbon sources of energy that could potentially replace fossil fuels with sufficient development and to propel the use of these sources. To do that barriers related to large-scale implementation need to be removed to build adequate confidence in low-carbon energy sources such as geothermal energy. Therefore, this thesis aims to reduce uncertainty that comes with geothermal energy implementation through a modeling approach. Enhanced Geothermal Systems (EGS) are an example of low-carbon energy resources with adequate potential to extract heat stored in the earth by creating subsurface fracture systems when hydrothermal fluids and permeability do not exist naturally. In EGS reservoirs, permeability is simulated hydraulically to ensure sufficient exposure of rock surfaces between injection and production wells. A well-established problem in EGS reservoir management is maldistribution of fluid flow in fractures that can impact reservoir lifetime, especially in fracture systems characterized by non-uniform permeabilities. Identification of such heterogeneity in permeability will reduce uncertainties in forecasting thermal performance. A key motivation for this study is to develop improved methods for predicting the long-term thermal and hydraulic performance of fractured geothermal reservoirs by incorporating methods from Machine Learning in reservoir and subsurface characterization.