Real-time estimation of geomechanical characteristics using drilling parameter data and LWD

In the pursuit of real-time estimation of geomechanical characteristics, this study integrates surface drilling telemetry with Logging While Drilling (LWD) to predict shear wave velocity (Vs) and other essential elastic properties of rock formations. Real-time prediction of these parameters is crucial for enhancing wellbore stability, fracture propagation, and geosteering operations, thereby improving both safety and operational efficiency. Traditional methods, which rely solely on conventional well-logging data, often fail to incorporate the dynamic information embedded within drilling mechanics, limiting their applicability in real-time decision-making.

Empirical validation using real drilling data from the Volve oil field demonstrated the enhanced performance of our self-attention-based Transformer model through the integration of drilling engineering parameters. In the initial testing, the model significantly improved the accuracy of predicting Vs, increasing it from 92% to 97.2%, alongside notable improvements in elastic property predictions. Specifically, the mean absolute error (MAE) for shear modulus decreased from 0.186 to 0.059, and bulk modulus from 0.189 to 0.040. Additionally, cross-validation using well F11A further confirmed the model's robustness, with the MAE for shear modulus decreasing from 0.134 to 0.053 upon incorporating drilling data. Compared to traditional LSTM-based models, the Transformer exhibited superior capability in extracting temporal features, validating its effectiveness in real-time elastic property prediction. These results underscore the model's capacity to enhance real-time decision-making in drilling operations.