A Cyber-Physical Approach to Early Kick Detection

Abstract

Abstract This paper describes a novel approach to early kick detection. A cyber-physical approach is utilized to improve the speed and consistency at which a kick can be identified during drilling operations, thus automating the kick detection process. The proposed methodology combines physics-based modelling with Bayesian mathematics for detecting subtle changes in noisy and uncertain measurements. For early kick detection purposes, the physics model is that of a lumped parameter model that describes the fluid flow in a well and the noisy and uncertain measurements are the data streams from rig sensors available during well construction operations, which are used to fit/correct the results of the model via an extended Kalman filter approach. A tool was built that is able to consume real-time and archived data, solve the lumped parameter model and Kalman filter implementation, and output the computed results for real-time viewing purposes. Historical datasets were used to demonstrate that the tool is able to detect kicks earlier than conventional methods, while providing an acceptable low false alarm rate. A real-time trial was conducted within a real-time monitoring center to evaluate the performance of the tool as an effective early kick detection technology via a set of key performance indicators and test the developed operational protocols and training materials with monitoring center personnel. The metrics put in place evaluated the tool in three categories: software robustness, data quality and early kick detection technical performance. A test plan was developed with test procedures and a workflow for utilizing the information given by the tool in the context of a real-time monitoring center. No kicks occurred during the construction of the well. However, certain operational events that resemble a kick demonstrated that the tool is able to identify them and trigger an alarm. The overall results of the real-time trial were favorable and potential enhancements to the tool were identified.