Prediction of Subsurface Geological-Induced Drilling Non-Productive Time Incidents Using Geophysical Inputs

Abstract:

A wellbore trajectory is a critical aspect of early well planning that requires the integration of cross-domain inputs. Conventionally, a wellbore trajectory focuses on anti-collision analysis and operational limits. In an effort towards optimization, the desired wellbore trajectory should deliver a safe path and minimize encountering subsurface geohazards and costs.

We propose a system to predict the probabilities of geological induced hazards in the wellbore trajectory and optimize its path during the planning phase of the well construction. The system implements a new dimension, i.e., drilling hazards seismic attributes (DHSA), to evaluate the proposed wellbore trajectory and quantify its associated risks. This process first aggregates cross-disciplinary data, such as geophysical seismic volumes, geological formation tops, the field historical drilling data (surface data and nonproductive time incidents), and the planned well specifications. The datasets undergo two main stages of data processing: spatial geophysical transformation and machine learning-based classification. The spatial geophysical transformation is dedicated to the geophysical data only (reflectivity and velocity volumes), where they go through an ensemble of transformations to produce relevant datasets that can be employed for detection and separation of subsurface geohazard zones. The underlining principle here is geophysical subsurface anomalies (velocity, density, uniaxial compressive strength, etc.) have a geophysical signature in the seismic volumes. The ensemble of transformations consists of [P-wave velocity to bulk density, porosity, and uniaxial compressive strength of rock], [seismic amplitude to 3D coherence attribute, and edge mapping], among others. The geophysical transformed data, historical drilling data, and geological data are aggregated and are the input to a machine learning imbalance classification algorithms (SMOTE oversampling and cost-sensitive support vector machines) aiming to predict a score representing the probability of drilling geohazards. The model is built as a depth post-stack window attribute, resulting in the DHSA cube. This DHSA attribute allows the evaluation of the drilling risks for a proposed trajectory and predicts the probabilities of geological induced NPT.