Implement intelligent dynamic analysis of bottom-hole pressure with naive Bayesian models

Abstract

During the drilling process, measured bottom-hole pressure data were prone to distortion and even no data are fed back, besides, the bottom-hole pressure calculation model could not reflect the live measured values. As a consequence, inaccurate bottom-hole pressure monitoring would bring enormous safety risks to drilling operations. Data mining was an advanced method used to sort out, discover and set up models from large relevant data sets. In the monitoring of bottom-hole pressures, it was necessary to conduct an effective and overall monitoring during the drilling process. Therefore, this paper proposed the k-means clustering method to optimize Naive Bayesian models in combination with the bottom-hole pressure monitoring theory, a k-means clustering optimized Naive Bayesian model for implementation of intelligent dynamic analysis of bottom-hole pressure was established. Such model could be utilized for correcting bottom-hole pressures calculated by the traditional hydraulic model, and then the corrections were taken for comparison with measured bottom-hole pressures so as to make calculations be of minimal errors. Field data were also taken for analysis, and the results suggested that the k-means based Naive Bayesian models for correction to calculated bottom-hole pressures had smaller deviations which fell within safe deviation monitoring range of drilling pressures, and could meet the requirements of normal drilling operation.