Trend Analysis for Calibrated Active System Volume and Differential Flow Significantly Reduces False Positive Kick Alarm During Drilling

Abstract Rapid kick detection is an essential part of rig safety. Identifying potential kick events in time is valuable and is an urgent and critical task for rig personnel and remote Operation Support Center(OSC) users. A new monitoring system has been developed by firstly calibrating the active system volume for pipe and cuttings displacement during drilling and then by detecting a statistically significant trend in the observed data rather than to rely solely on absolute values and threshold crossing to maximize the sensitivity whilst minimizing the number of false alarms. More than hundreds of historical kick alarms were reviewed and the corresponding mud logging data before the events occurred were displayed in log format against time for clear identification of anomalies in three different tracks: one track for pit volume, one for flow-in and flow-out, the third hole depth, bit depth and block position. By applying an automated rule-based system, the states of rig activities are also displayed simultaneously by detection of bit movement, pump, rotation status and other elements. It is noticed that the data absolute value, data noise, and rig operations such as sudden string movements and flow transients has significant interference to the detection of kick event and that the differential flow is more sensitive early kick indicator than pit volume in most of the historical cases. A pit volume calibrated for drill pipe and cuttings displacement during drilling and a trend analysis by applying Page-Hinkley method for monitoring both active system volumes and differential flows are established and the monitoring system is updated. The system has been tested and refined with the data flows of the historical cases in labs to maximize the sensitivity whilst minimizing the number of false alarms. A significant reduction of false positive alarms is achieved by reducing triggering of 1-2 alarms per well per week while maintaining a 100% true positive identification rate by comparing 1-2 alarms per day per well in the old system which was relied solely on absolute values and threshold crossing. This paper summarizes the current methods for the detection of kick event, the study and the findings from replay of historical kick alarm cases, the methods of calibrating pit volume during drilling operation, the algorithms for rig state detection, and the application of early kick detection methods of differential flow and flow-back fingerprint.