The Geothermal Artificial Intelligence for geothermal exploration

Introduction to Geothermal Exploration

Geothermal Exploration: Geothermal exploration involves a comprehensive analysis and strategic management of uncertainties associated with geothermal resources. These uncertainties stem from various factors such as geological complexities, the availability of accurate data, and the unpredictability of geothermal systems, making investment and operational decisions challenging for stakeholders in the energy sector.
Technological Integration: The integration of advanced technologies such as Remote Sensing (RS), Machine Learning (ML), and Artificial Intelligence (AI) is increasingly being recognized as an essential strategy to improve the effectiveness and efficiency of geothermal exploration. By leveraging these methodologies, energy developers can enhance their understanding of geothermal reservoirs and reduce uncertainties in the exploration process.

Technological Integration in Geothermal Exploration

Technological integration plays a crucial role in enhancing geothermal exploration through the use of advanced methodologies like Remote Sensing (RS), Machine Learning (ML), and Artificial Intelligence (AI). Each of these technologies contributes distinct advantages by either improving data collection processes or refining analysis and decision-making:

Remote Sensing (RS): This method facilitates the collection of comprehensive data without direct contact with the geothermal site. Remote Sensing technologies utilize satellite imagery and aerial surveys to assess surface conditions, identify thermal anomalies, and understand geological formations that suggest geothermal potential. Tools such as LANDSAT satellites and thermal infrared sensors allow for the identification of surface temperature variations that may indicate underlying geothermal systems.
Machine Learning (ML): Machine Learning techniques enable the processing of vast datasets to identify patterns and correlations related to geothermal activity. By employing algorithms that learn from historical and current geological data, stakeholders can make more informed predictions regarding the presence and viability of geothermal resources.
Artificial Intelligence (AI): AI encompasses sophisticated analytical frameworks that enhance decision-making processes in geothermal exploration. By simulating human cognitive functions and automating data evaluation, AI helps optimize resource management efforts and improve predictive modeling for locating geothermal reserves, ultimately leading to more successful exploration initiatives.

The cumulative use of RS, ML, and AI not only streamlines the geothermal exploration process but also significantly enhances the precision of resource assessments, paving the way for sustainable energy solutions in the future.

Remote Sensing (RS) Remote sensing allows for the collection of data from satellites or aerial platforms without direct contact with the ground. This technology is particularly beneficial in geothermal exploration as it provides extensive and detailed information about surface temperatures, geological features, and potential geothermal activity.

Data Collection: RS utilizes various tools such as LANDSAT satellites, aerial thermal imaging, and hyperspectral sensors to capture high-resolution data.
Benefits: Offers a cost-effective and efficient means of assessing large areas that may be challenging to access directly. Enables the identification of thermal anomalies and fault lines which are indicators of geothermal potential.

Machine Learning (ML) Machine learning, a subset of AI, refers to the computational methods that enable computers to learn from and make predictions based on data. In geothermal exploration, ML algorithms analyze complex datasets to identify patterns and predict the likelihood of geothermal energy presence.

Predictive Analysis: ML models process historical data along with current geological surveys to enhance the accuracy of geothermal potential assessments.
Data Handling: Methods like supervised learning, where labeled datasets train models, and unsupervised learning, where the model identifies patterns without prior labels, are critical in differentiating geothermal sites from non-geothermal ones.

Artificial Intelligence (AI) Artificial intelligence encompasses ML and further extends to incorporating theoretical frameworks mimicking human cognitive functions. AI applications in geothermal exploration are aimed at optimizing resource management and improving exploration outcomes.

Geothermal AI Development: Utilizing deep learning algorithms tailored for geothermal data, AI can enhance the predictive capacity for geothermal resources. AI systems can learn from historical site data to improve their performance in recognizing new potential sites.
Integration of Multiple Data Types: AI systems synthesize various data forms—such as geological, geophysical, and geochemical data—allowing for a more holistic view of geothermal potential.

Conclusion The integration of RS, ML, and AI not only streamlines data collection and analysis in geothermal exploration but also significantly enhances the precision of resource assessments. This combination of technologies is essential for overcoming the existing challenges in geothermal energy development, paving the way for sustainable energy solutions.</li><li>Indicators of Geothermal Potential: Key indicators include mineral markers, surface temperature, faults, and geological deformation. These help assess geothermal potential.</li></ul><h4 id="c6b39e47-0996-4e2b-9b1e-5e36e55f86f0" data-toc-id="c6b39e47-0996-4e2b-9b1e-5e36e55f86f0" collapsed="false" seolevelmigrated="true">Challenges in Geothermal Energy Adoption</h4><ul><li>Energy Contribution: In 2019, geothermal energy comprised only 0.19% of installed electricity capacity and contributed 0.33% of total electricity generation across the globe. This limited presence in the energy market highlights the underutilization of geothermal resources compared to more dominant energy sources like fossil fuels, wind, and solar. Despite its potential for providing a stable and reliable energy source, geothermal still faces significant hurdles in expanding its role in the energy mix.</li><li>Investment Barriers: Several factors impede the growth of geothermal energy investment, including:<ul><li>Resource Availability: The geographical distribution of geothermal resources is uneven, with many regions lacking suitable sites for development. This limitation can restrict potential investors from pursuing geothermal projects in certain areas.</li><li>Investment Risk: High upfront capital costs and the uncertainty surrounding resource viability contribute to perceived investment risks. Investors may be hesitant to fund geothermal projects due to the long timelines needed for exploration and development before any return on investment can be realized.</li><li>Understanding of the Technology: Many potential stakeholders, including investors and local communities, may lack a comprehensive understanding of geothermal technologies and their benefits. This knowledge gap can lead to reluctance in supporting geothermal initiatives.</li></ul></li><li>Economic Evaluation: The Levelized Cost of Energy (LCOE) methodology is commonly utilized to evaluate the costs associated with various energy generation projects over their operational lifespans. However, the traditional use of LCOE for geothermal projects often overlooks important factors such as:<ul><li>Risk Effects: Selection bias can distort the actual economic assessments, leading to an underestimation of the risks associated with geothermal investments, such as drilling challenges or fluctuating resource availability.</li><li>External Costs: Environmental impacts and social considerations, which could influence overall project viability and public acceptance, are commonly excluded from traditional LCOE analyses.</li></ul></li><li>Risk Mitigation Strategies: Proper risk management is essential for the successful adoption of geothermal energy. Some effective strategies include:<ul><li>Phased Exploration Programs: Gradually investing in exploration and development allows for risk assessment and adjustments based on preliminary findings. This approach enables investors to commit incrementally, reducing their exposure to financial losses.</li><li>Public-Private Partnerships: Collaborative efforts between governmental bodies and private companies can help share risks and develop geothermal resources more effectively. Initiatives such as grants and subsidies can also incentivize investment in geothermal projects.</li><li>Advanced Technologies: Leveraging advanced exploration techniques, like machine learning and artificial intelligence, can increase the likelihood of successful resource identification and reduce drilling risks, making geothermal projects more attractive to investors.</li></ul></li></ul><ul><li>Investment Barriers: Factors such as resource availability, investment risk, and understanding of the technology hinder geothermal energy adoption.</li><li>Economic Evaluation: Levelized Cost of Energy (LCOE) methodology helps assess costs over the life of projects but often neglects risk effects because of selection bias.</li><li>Risk Mitigation Strategies: Strategies like phased exploration programs are essential for maximizing return on investment and managing risks.</li></ul><h3 id="a1d88720-7cb0-43c3-a0f4-37074169d455" data-toc-id="a1d88720-7cb0-43c3-a0f4-37074169d455" collapsed="false" seolevelmigrated="true">Exploration Uncertainties</h3><ul><li>Types of Uncertainties:<ul><li>Geological Uncertainty: Results from incomplete knowledge of subsurface characteristics affecting production strategies.</li><li>Technological Uncertainty: Arises from limited data types, statistical analysis limits, and measurement errors.</li><li>Economic Uncertainty: Related to changes in energy technology, forecasting difficulties, and competition impacts.</li></ul></li></ul><h3 id="a99e4d2d-6d6b-4278-a27e-5a3e6da353d2" data-toc-id="a99e4d2d-6d6b-4278-a27e-5a3e6da353d2" collapsed="false" seolevelmigrated="true">Methodology Overview</h3><h4 id="47111aee-1e50-4b38-bdd6-3b07feb33809" data-toc-id="47111aee-1e50-4b38-bdd6-3b07feb33809" collapsed="false" seolevelmigrated="true">Proposed Integrated Methodology</h4><ul><li>Combines RS, ML, and AI to create an initial geothermal potential assessment.</li><li>Tested in two geothermal sites: Brady (with surface manifestations) and Desert Peak (blind site).</li><li>Satellite Data & Geospatial Analysis: Various data forms were processed to identify patterns related to surface manifestations.</li><li>Development of Geothermal AI to predict geothermal potential based on identified patterns.</li></ul><h4 id="e592bd4b-901b-4ad8-b11b-210809ffb339" data-toc-id="e592bd4b-901b-4ad8-b11b-210809ffb339" collapsed="false" seolevelmigrated="true">Methodological Stages</h4><ol><li>Selection of Inputs: Choosing geothermal indicators for the AI model.</li><li>Geothermal Indicators Analysis: Performing spatial and temporal analyses to extract data patterns enhancing AI learning.</li><li>Automatic Labeling: Utilizing unsupervised ML to classify geothermal and non-geothermal sites for AI training.</li><li>Geothermal AI Development: Crafting an AI that employs a deep learning algorithm tailored for geothermal data through spatial statistics.</li><li>Accuracy Assessment: Evaluating the AI's performance in predicting geothermal sites based on independent datasets.</li></ol><h3 id="ac72dfac-07be-41e0-9665-4228c9554d71" data-toc-id="ac72dfac-07be-41e0-9665-4228c9554d71" collapsed="false" seolevelmigrated="true">Indicator Analysis**</h3><h4 id="18c9c2a4-3669-4153-a9bb-0419d806ca0c" data-toc-id="18c9c2a4-3669-4153-a9bb-0419d806ca0c" collapsed="false" seolevelmigrated="true">Types of Geothermal Indicators</h4><ol><li>Temperature: Analysis of LANDSAT thermal images to identify hot zones.</li><li>Faults: Compiling fault maps and converting to density maps for influence zone integration in Geothermal AI.</li><li>Mineral Markers: Identifying hydrological alterations in mineral compositions that indicate geothermal activity.</li><li>Deformation: Analyzing subsidence and uplift effects on geothermal systems through Persistent Scatterer Interferometric Synthetic Aperture Radar (PSInSAR).</li></ol><h4 id="8d2d33a8-03f6-49e3-abac-59d49dccf5c3" data-toc-id="8d2d33a8-03f6-49e3-abac-59d49dccf5c3" collapsed="false" seolevelmigrated="true">Data Usage and Sources</h4>A thorough understanding of the selected geothermal indicators is critical for effective exploration and potential resource development. Below is a detailed summary of the chosen indicators, including their data sources and resolutions:<ul><li>Temperature: Utilization of LANDSAT surface temperature data, which employs an advanced thermal infrared sensor. This data is pivotal in identifying hot zones that may indicate geothermal activity. The resolution of 30 meters allows for precise mapping of surface temperatures, facilitating the detection of thermal anomalies associated with geological features.</li><li>Faults: The Nevada Geological Survey fault maps provide crucial information about the structural geology within geothermal fields. These maps are essential for analyzing the spatial distribution of fault lines, which play a significant role in geothermal energy potential. By converting these maps into density maps, researchers can better assess the influence of faults on geothermal systems and their correlation to geothermal reservoirs.</li><li>Mineral Markers: The use of HyMap hyperspectral imagery at 3-meter resolution enables the identification and analysis of hydrological alterations in mineral compositions that signal geothermal activity. This high-resolution data allows for detailed mapping of surface minerals, essential for interpreting the geothermal signatures that indicate potential resource locations.</li><li>Deformation: The Synthetic Aperture Radar (SAR) data from Sentinel-1 provides valuable insights into ground deformation, a major indicator of subsurface geothermal activity. This radar technology contributes to Persistent Scatterer Interferometric Synthetic Aperture Radar (PSInSAR) analyses, allowing for the monitoring of subsidence or uplift events over time, which are critical for understanding the dynamics of geothermal systems.</li></ul><h4 id="551f6eda-8100-4df7-9f37-2e96099fab86" data-toc-id="551f6eda-8100-4df7-9f37-2e96099fab86" collapsed="false" seolevelmigrated="true">Automatic Labeling Process</h4><ul><li>Complex Nature: The automatic labeling process for identifying geothermal from non-geothermal areas is inherently complex due to the intricate and heterogeneous nature of the data involved. Expert knowledge is often necessary to interpret and analyze the data accurately; however, the impracticality arises from the sheer volume of data that needs to be processed, making it difficult to rely solely on human expertise. This situation necessitates the development of robust algorithms that can automate the classification process while minimizing reliance on manual input.</li><li>Unsupervised Learning Approach: To tackle these challenges, an unsupervised learning approach is utilized, specifically through the implementation of a self-organizing map (SOM) algorithm. This type of neural network is effective in mapping high-dimensional datasets into lower-dimensional (often two-dimensional) representations while preserving the topological relationships within the data. By delineating geothermal versus non-geothermal areas, the SOM algorithm identifies patterns and clusters within the data without the need for pre-labeled examples. This allows for the automatic identification of areas with geothermal potential based on calculated similarity and distance metrics between various data points. The result is a more efficient and scalable classification process, paving the way for enhanced geothermal exploration efforts and reducing the time and resources required for initial assessments.</li></ul><h3 id="196bde83-974c-40f8-9e55-4e7755feacc7" data-toc-id="196bde83-974c-40f8-9e55-4e7755feacc7" collapsed="false" seolevelmigrated="true">Development of Geothermal AI</h3><ul><li>Machine Learning Models: Various methods were utilized in the context of spatial correlations and understanding data characteristics.</li><li>CNN Architecture: Modifications included kernel designs based on autocorrelation analysis to enhance prediction accuracy. #### Machine Learning Models - Overview of Methods: In the development of Geothermal AI, various machine learning models were employed to improve the understanding of spatial correlations and to analyze the characteristics of geothermal data. Different algorithms were explored to identify the best-performing models for predicting potential geothermal sites. Among these methods, supervised and unsupervised learning techniques were utilized, allowing for a comprehensive approach to data interpretation and validation. - Spatial Correlation Analysis: Special focus was placed on understanding the spatial relationships among the geothermal indicators collected from various sources. By analyzing these correlations, the models were able to identify patterns that are indicative of geothermal activity, leading to more accurate predictions of geothermal resource locations. - Feature Engineering: The process of feature engineering was critical in enhancing the performance of the AI models. This involved the selection and transformation of relevant features from the dataset—such as temperature readings, fault lines, and mineral compositions—which significantly affect the models' predictive capabilities. #### CNN Architecture - Convolutional Neural Networks (CNNs): The architecture of CNNs was employed due to their effectiveness in image recognizing tasks, which is crucial for analyzing thermal and geological imagery related to geothermal exploration. - Kernel Designs and Autocorrelation Analysis: Modifications to the CNN architecture included the development of specialized kernel designs informed by autocorrelation analysis. This technique allowed for the identification of patterns within the data that may not be visible with traditional methods, hence enhancing the model's ability to learn from historical geothermal data accurately. - Layer Adjustments: The CNN layers were optimized for better performance by adjusting parameters such as filter size and stride length, which tailored the model for the specific spatial characteristics found in geothermal datasets. This refinement was essential in ensuring that the models could efficiently process the high-dimensional data typical of remote sensing inputs. #### Model Training and Validation - Training Process: The training of the CNN involved curating a robust dataset that included both geothermal and non-geothermal sites, allowing the model to learn distinguishing characteristics between potential geothermal locations and other land types. - Validation Framework: A validation framework was established to assess the model's performance continuously. Performance metrics such as accuracy, precision, and recall were used to monitor how well the model generalizes across different geothermal sites, ensuring reliability in predictions before real-world application.</li></ul><h3 id="4c658697-9660-4701-aeb7-ebe79b6fb642" data-toc-id="4c658697-9660-4701-aeb7-ebe79b6fb642" collapsed="false" seolevelmigrated="true">Training and Testing of Geothermal AI**</h3><ul><li>Model Evaluation: Trained on Brady data (with surface manifestations) and tested across independent sites (including Desert Peak).</li><li>Performance Metrics: Achieved accuracy levels between 92% and 95%, even with the blind site.</li><li>Independent Testing Success: Models trained on one site were effectively applied to the other, showcasing generalization capabilities.</li></ul><h3 id="597ff0a4-b499-4d23-b6bd-e5ea211c9c70" data-toc-id="597ff0a4-b499-4d23-b6bd-e5ea211c9c70" collapsed="false" seolevelmigrated="true">Conclusion and Future Directions</h3><ul><li>Emphasizes integrating RS, ML, and AI for better geothermal exploration outcomes.</li><li>Suggests refining models with more data, enhancing labels, and incorporating subsurface assessments for comprehensive geothermal resource evaluation.</li></ul>