Session: 01-01: AI for Energy Sustainability I

Paper Number: 129901

129901 - Enhancing Ai-Driven Co2 Plume Geothermal Power Production – a Pohokura Field Benchmark Analysis 

Abstract: 

There has been significant interest in utilizing CO2 for power generation. Because CO2-plume geothermal (CPG) renewable energy may both enhance CO2 utilization and provide energy security, there has been a lot of interest in research on this technology. CO2 plume geothermal generates geothermal energy by using CO2 that is circulated in saline aquifers and other subsurface reservoir types. We performed reservoir simulations of coupled transport of formation fluid, injected CO2, and heat in a heterogeneous 3D reservoir model to quantify recovered thermal energy via CPG. In order to simulate the physical simulation, we provide an integrated simulation and deep learning optimization approach in this research. By utilizing such an approach, we can speed up decision-making for the best possible power generation from CPG while lowering computational costs. We introduce a novel deep-learning optimization methodology designed to maximize the CPG system's power generation. A CO2 injection production data-driven simulation model of four wells within the Taranaki basin, New Zealand, was built based on a geological reservoir model of the Pohokura gas field. The outcomes are integrated into the deep learning framework as input data. Based on a simulation framework, the system optimizes electricity generation through ensemble-bagging. In order to retain the model's interpretability while modeling the CO2 plume generation from the reservoir over time, the bootstrap aggregation uses decision trees' advantages. We tested this novel framework using a simulated CO2 storage in the Taranaki basin's Pohokura gas field, which has been extensively researched for CO2 storage because it has a sizable saline aquifer that might be useful for producing CPG-derived energy and for CO2 storage purposes. We simulated twenty years of CO2 production and injection for geothermal energy generation. The network demonstrated high training performance, and the model's effectiveness was assessed during the next three years of energy production. After that, the deep-learning framework is part of a global optimization framework to adjust CO2 injection and power generation from the individual CPG stations to maximize energy production and the total carbon footprint. With its novel approach to improving energy output from CO2 reservoirs, the new deep-learning ensemble-bagging optimization framework for CPG power generation offers a viable means of reducing carbon footprint while producing electricity. The framework's adaptability to varied input parameters and range of forecastable time series are its main features. For CPG to effectively capture changes in the temperature responses and temporal dynamics among the several CO2 injection and production wells, this is especially crucial.

Presenting Author: Abdulaziz Qasim Saudi Aramco

Presenting Author Biography: Abdulaziz is an expert in sustainability and CCUS and leader of the CO2 project.

Authors:

Klemens Katterbauer Saudi Aramco
Abdallah Shehri Saudi Aramco
Abdulaziz Qasim Saudi Aramco
Ali Yousef Saudi Aramco