The success of machine learning (ML) techniques implemented in different industries heavily rely on operator expertise and domain knowledge, which is used in manually choosing an algorithm and setting up the specific algorithm parameters for a problem. Due to the manual nature of model selection and parameter tuning, it is impossible to quantify or evaluate the quality of this manual process, which in turn limits the ability to perform comparison studies between different algorithms. In this study, we propose a new hybrid approach for developing machine learning workflows to help automated algorithm selection and hyperparameter optimization. The proposed approach provides a robust, reproducible, and unbiased workflow that can be quantified and validated using different scoring metrics. We have used the most common workflows implemented in the application of artificial intelligence (AI) and ML in engineering problems including grid/random search, Bayesian search and optimization, genetic programming, and compared that with our new hybrid approach that includes the integration of Tree-based Pipeline Optimization Tool (TPOT) and Bayesian optimization. The performance of each workflow is quantified using different scoring metrics such as Pearson correlation (i.e., R² correlation) and Mean Square Error (i.e., MSE). For this purpose, actual field data obtained from 1567 gas wells in Marcellus Shale, with 121 features from reservoir, drilling, completion, stimulation, and operation is tested using different proposed workflows. A proposed new hybrid workflow is then used to evaluate the type well used for evaluation of Marcellus shale gas production. In conclusion, our automated hybrid approach showed significant improvement in comparison to other proposed workflows using both scoring matrices. The new hybrid approach provides a practical tool that supports the automated model and hyperparameter selection, which is tested using real field data that can be implemented in solving different engineering problems using artificial intelligence and machine learning. The new hybrid model is tested in a real field and compared with conventional type wells developed by field engineers. It is found that the type well of the field is very close to P50 predictions of the field, which shows great success in the completion design of the field performed by field engineers. It also shows that the field average production could have been improved by 8% if shorter cluster spacing and higher proppant loading per cluster were used during the frac jobs.

Major concerns of carbon dioxide (CO₂) sequestration in subsurface formations are knowledge of the well injectivity and gas storage capacity of the formation, the CO₂ pressure and saturation plume extensions during and after injection, and the risks associated with CO₂ leakage and fault reactivation. Saline reservoirs are considered as one of the target formations for CO₂ sequestration through structural, residual, dissolution, and mineral trapping mechanisms. The boundary condition of the saline reservoir dictates the pressure and saturation plume extension of the injected supercritical CO₂ that could expand over large distances. This can lead to sources of risk, e.g., leakage and/or fault reactivation due to presence of wells, thief zones, and geological discontinuities. Therefore, there is a critical need to develop a model that describes how risk-related performance metrics (i.e., the CO₂ saturation plume size, the pressure differential plume area, and the pressure differential at specific locations) vary as a function of the size of injection, time following injection, injection operations, and geologic environment. In this study, a systematic reservoir modeling studies of anthropogenic CO₂ sequestration in Citronelle dome, Alabama, was performed where all relevant scenarios and conditions to address the questions of the saturation and pressure plume size in the area of review (AoR) and post-injection site care (PISC) are considered. The objective for this study was firstly to systematically simulate CO₂ sequestration, i.e., saturation dynamics, and pressure behavior over a range of operational and geological conditions and to derive conclusions about the factors influencing saturation and pressure plume size, post-injection behavior, and the risk associated with them, by developing third-generation reduced order models (ROMs) for reservoir behavior. Finally, to assess the uncertainty associated with our studies, Latin Hypercube Sampling (LHS) together with an experimental design technique, i.e., Plackett–Burman design, was used. Application of Pareto charts and respond surfaces enabled us to determine the most important parameters impacting saturation and pressure plume sizes and to quantify the auto- and cross-correlation among different parameters in both history-matched and upscaled models.

Summary: This paper presents a novel workflow for high-precision leak detection in pipeline networks using the negative pressure wave (NPW) technique. The proposed workflow addresses challenges associated with noisy and convoluted pressure transducer data, rapid pressure decay, and the need for robustness in leak event detection. To overcome these challenges, the workflow incorporates data preprocessing techniques for cleansing, normalization, and denoising, as well as dynamic pressure control limit lines to differentiate between pump and leak events. Multiple transducer analysis techniques are used to minimize false positives. Synthetic leak scenarios are generated using the Water Network Tool for Resilience (WNTR) package, enabling a comprehensive assessment of the workflow’s performance. The generated scenarios are validated through pressure history matching against field inline pressure recordings. A dashboard is developed for real-time visualization and verification of leak events. The effectiveness of the workflow is demonstrated through testing on a real network, resulting in the successful detection and precise localization of a confirmed leak event. The workflow proves its capability to achieve high accuracy, with a 100-m resolution in a complex network configuration with 29 pipe sections and 1-Hz pressure signal recordings. For synthetic leak events, a 10-Hz pressure signal is utilized, achieving a remarkable 10-m accuracy. Moreover, the integration of the workflow with supervisory control and data acquisition (SCADA) systems is showcased, highlighting its potential for near real-time leak detection in practical applications. Overall, this paper presents a comprehensive and effective workflow for high-precision leak detection and localization in pipeline networks, offering valuable insights into improving the efficiency and reliability of leak detection systems.

Optimizing unconventional field development requires simultaneous optimization of well spacing and completion design. However, the conventional practice of using cross plots and sensitivity analysis via Monte Carlo simulations for independent optimization of well spacing and completion design has proved inadequate for unconventional reservoirs. This is due to the inability of cross plots to capture non-linear cross-correlations between parameters affecting hydrocarbon production, and the computational expense and difficulty of Monte Carlo simulations. Recently, automated machine learning (AutoML) workflows have been used to tackle complex problems. However, applying AutoML workflows to engineering problems presents unique challenges, as achieving high accuracy in forecasting the physics of the problem is crucial. To address this issue, a new physics-informed AutoML workflow based on the TPOT open-source tool developed that guarantees the physical plausibility of the optimum model while minimizing human bias and uncertainty. The workflow has been implemented in a Marcellus Shale reservoir with over 1500 wells to determine the optimal well spacing and completion design parameters for both the field and each well. The results show that using a shorter stage length and a higher sand-to-water ratio is preferable for this field, as it can increase cumulative gas production by up to 8%. Additionally, it is observed that fifty-percentile cumulative gas predictions are in close agreement with actual field productions. Thematic collection: This article is part of the Digitally enabled geoscience workflows: unlocking the power of our data collection available at: https://www.lyellcollection.org/topic/collections/digitally-enabled-geoscience-workflows

This study utilizes machine learning to quantify CO₂ plume extents by analyzing microseismic data from the Illinois Basin Decatur Project (IBDP). Leveraging a unique dataset of well logs, microseismic records, and CO₂ injection metrics, this work aims to predict the temporal evolution of subsurface CO₂ saturation plumes. The findings illustrate that machine learning can predict plume dynamics, revealing vertical clustering of microseismic events over distinct time periods within certain proximities to the injection well, consistent with an invasion percolation model. The buoyant CO₂ plume partially trapped within sandstone intervals periodically breaches localized barriers or baffles, which act as leaky seals and impede vertical migration until buoyancy overcomes gravity and capillary forces, leading to breakthroughs along vertical zones of weakness. Between different unsupervised clustering techniques, K-Means and DBSCAN were applied and analyzed in detail, where K-means outperformed DBSCAN in this specific study by indicating the combination of the highest Silhouette Score and the lowest Davies–Bouldin Index. The predictive capability of machine learning models in quantifying CO₂ saturation plume extension is significant for real-time monitoring and management of CO₂ sequestration sites. The models exhibit high accuracy, validated against physical models and injection data from the IBDP, reinforcing the viability of CO₂ geological sequestration as a climate change mitigation strategy and enhancing advanced tools for safe management of these operations.