Mitigating Risks in Well Construction Through Real-Time Reverse Multiwell Analysis to Enhance Updating the Drilling Window

Abstract This paper presents a new multiwell workflow that uses stress inversion and machine learning algorithms to update a field-scale 1D mechanical earth model (1D MEM) and wellbore stability (WBS) calculation in real time. The workflow addresses the common issue of discrepancies between newly acquired data and the predrill model. The model and drilling performance capabilities are compared with previously published real-time workflows based on the traditional single-well approach. The proposed reverse multiwell geomechanics workflow is a real-time solution that recycles information. Initially, the workflow updates a predrill database by connecting a new event to customized metrics such as location, geology, and risk classification. The updated database enhances data correlation and quality control by identifying field-scale patterns. The validated inputs are then used to recalibrate the entire model, including the offset-well predrill model, using stress inversion algorithms for poroelastic parameters and machine learning for log prediction. The output is automatically validated across all wells in the project to ensure consistency and predictability for the upcoming drilling operation. This paper demonstrates the core value of real-time drilling geomechanics in relation to its ability to handle uncertainties and data divergences. The current advanced computational hardware and software capabilities enable the processing of multiple wells in a single run using consistent parameters in a feasible time frame for real-time projects. Hence, this paper presents new tools for the geomechanics analyst to assess whether newly acquired data are outside predefined boundaries and whether recalibration of the model is necessary. The field-scale-validated model is used to revise the forecast simulation for upcoming drilling operations. In conclusion, real-time reverse multiwell analysis minimizes uncertainties by implementing a fully calibrated geomechanics model consistent with other wells in the field, leading to improved forecasts during drilling operations and well construction. A structured drilling-geomechanics database and an automated workflow using stress inversion and machine learning algorithms aim to improve the predictive capabilities of real-time drilling geomechanics through reverse multiwell analysis. This approach reduces uncertainties from the initial model, facilitates quick model recalibration, accelerates learning, and ensures the updates to the drilling window are consistent with field-scale data. Consequently, unexpected deviations in geology and rock properties can be seamlessly incorporated into the model, reducing nonproductive drilling time.