The upstream oil and gas industry has seen its share of innovation over the past several decades. The driving force behind these changes has always been a relentless push toward operational and capital efficiency. A breakthrough patented pressure diagnostic technique using offset sealed wellbores as monitoring sources was introduced at the 2020 Hydraulic Fracturing Technology Conference (Haustveit et al. 2020). This technique quantifies various hydraulic fracture parameters using only a surface gauge mounted on the sealed wellbore. The authors successfully automated the Sealed Wellbore Pressure Monitoring (SWPM) analysis procedure using a cloud-based analytical platform (CBAP) designed to ingest, process, and analyze high-frequency hydraulic fracturing data (Iriarte et al. 2021b). The minimum data for the analysis consists of the standard frac treatment data combined with the high-resolution pressure gauge data for each sealed wellbore. The team developed machine learning algorithms to identify the key events required by a sealed wellbore pressure analysis: the start, end, and magnitude of each pressure response detected in the sealed wellbore while actively fracturing offset wells. The result is a rapid, repeatable SWPM analysis that minimizes individual interpretation biases. Since then, over 10,000 stages have been analyzed with SWPM in every major North and South American unconventional basin.
The next logical step in the process was to move from post-treatment to real-time analyses. This required an extensive data set to train the real-time models. The training data set includes two types of data: active well data including treating pressures and slurry rates for 1000+ stages from all major North American basins; and 2500+ hours of monitoring well pressure and temperature data streams. The authors use signal processing techniques to mitigate noise, easily accommodate business rules, and follow the subject matter experts’ decision logic. The data is combined with high-resolution pressure gauge data. Machine learning algorithms were developed to identify the start, end, and magnitude of each pressure response detected in the sealed wellbores while actively fracturing offset wells. The model updates its predictions as new data are collected, generating predictions every few seconds on average. The length of the streaming window analyzed by the model and the frequency of the analysis can be modified to accommodate a variety of internet and streaming conditions. This approach provides a robust, automated, and extremely performant model that easily accommodates operating constraints.
Real-time cloud-based streaming paired with machine learning allows much easier decision making on-the-fly. Moreover, the proposed methods are designed so that real‐time updates can be done efficiently. One of the benefits of real-time data is the ability to manage by exception. Using alerts that are triggered by customizable thresholds, remote engineers can be aware of any operational issues. Closely evaluating sealed wellbore pressure responses to changing completion designs in an active well allows further optimization of the completion process along with creating opportunities for saving costs.