Predicting Seismic-Based Anisotropy for Prevent Pre-Drill Risk Using a Novel Type Neural Network

Abstract

While drilling into fracture zones, lost circulation frequently occurs, resulting in a waste of productive operation severe cases, the well's destruction. However, due to complex development mechanisms and high heterogeneity, identifying and predicting fractures is extremely difficult. This study proposes a new drilling loss prevention idea to evaluate fractured lost circulation risk using seismic and wellbore data by a novel neural network. The approach works in two steps. First, the fracture anisotropy of a lost circulation sample curve is computed and interpreted using well logs. Second, using seismic attributes as constraints, a novel neural network is used to develop a prediction model. The field application in the Sichuan basin verifies the method's efficacy and confirms the method's ability for predicting lost circulation probability both along the well trajectory and in regions away from the drilled wells.