Machine Learning Based Techno-economic Assessment and Optimization of an Enhanced Geothermal System

Date
2024-06-21
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In the emergency to achieve decarbonization goals, transitioning from traditional fossil fuels to renewable energy sources is paramount within the energy sector. Geothermal energy, particularly through the utilization of Enhanced Geothermal Systems (EGS), is recognized as a promising low-carbon alternative for future energy supply, offering lower carbon intensity in electricity production. Despite its potential, conventional EGS operations face significant technical and economic challenges, compounded by the lack of an accurate and efficient optimization tool to enhance EGS profitability. This thesis addresses these challenges through three comprehensive studies. The first study introduces a novel CO2-water-EGS framework, applied to the Qiabuqia geothermal field in China, to explore the technical feasibility of overcoming the limitations of traditional EGSs (water-EGS and CO2-EGS). An integrated analysis, encompassing heat extraction and carbon storage, evaluates the impacts of seven operational factors. Key findings in this part are: (1) CO2-water-EGSs can recover more geothermal energy than water-EGSs and CO2-EGSs, whereas CO2-EGSs can store the largest volume of CO2 in the reservoir; (2) horizontal-well-EGSs generally offer superior heat mining capabilities over vertical-well-EGSs inspite of increased thermal breakthrough risks; (3) investigated technical factors show their complex correlations with technical performance, thereby necessitating the development of a high-performance optimization tool to maximize profitability. Secondly, an evaluation framework from technical and economic perspectives, incorporating the impact of carbon credit, is introduced for the first time to compare six EGS configurations. Sensitivity analyses explore the influence of various technical and economic variables. Results indicate that: (1) NPV (net present value) is a more effective metric than LCOE (levelized cost of electricity) for evaluating EGS economic viability; (2) CO2-water-EGSs are the most profitable options among all cases, with CO2-EGS showing an undesirable NPV due to excessive CO2 usage; (3) vertical-well-EGSs are economically superior to their horizontal counterparts; (4) economic outcomes are predominantly influenced by carbon credit rates, electricity market prices, and CO2 purchase prices, alongside the differential impacts of seven technical parameters. The third study develops an optimization framework incorporating machine learning and Differential Evolution (DE) algorithms, using NPV as the economic indicator for the Qiabuqia EGS. An optimal surrogate model determined through a comprehensive comparison of four machine learning algorithms, including Support Vector Machine (SVM), Extreme Gradient Boosting (XGB), and Artificial Neural Network (ANN), are integrated with a DE-based optimization process to identify a strategy yielding optimal NPV under operational constraints. Results show: (1) the ANN is more recommended to generate a surrogate model for the Qiabuqia EGS, which demonstrates a promising prediction accuracy with a R^2 value of 97.3%; (2) the ANN-based DE optimization method identifies an operational strategy resulting in the highest NPV of 39.8 M$, surpassing over 3,000 randomly generated numerical cases; (3) this optimization tool significantly reduces computational time, illustrating over a 100,000 times decrease compared to conventional numerical simulation.
Description
Keywords
Enhanced Geothermal System, Technical and Economic Assessment
Citation
Xue, Z. (2024). Machine learning based techno-economic assessment and optimization of an enhanced geothermal system (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.