LAUSR

Abstract Pore pressure prediction represents an important safety aspect of drilling engineering. Accurate pore pressure prediction is required for appropriate mud weight usage. Kick can occur when mud weight is lower than pore pressure gradient and this can result in disastrous events such as blowout when the kick is not properly controlled. Likewise, too high mud density can fracture the reservoir which can lead to several problems. Thus, the need to research on means of improving accurate pore pressure prediction during drilling is in order. In this article, two methodologies are presented. One of the methodologies is developed to utilize resistivity data for pore pressure prediction, and the other methodology is developed if resistivity and porosity data are available for pore pressure prediction. Several methodologies already exist for pore pressure prediction with resistivity data. Therefore, the methodology presented in this article is compared to other resistivity-based methodologies in order to observe their pore pressure prediction capabilities. Field data is used for testing prediction performance in terms of mean absolute percentage error, root mean square error and Pearson product moment correlation coefficient. Results of the test show that the methodology developed in this article performed best. Different logging/measurement parameters are used for pore pressure prediction e.g. resistivity log, sonic velocity, corrected d-exponent, etc. One way to improve accuracy of pore pressure prediction is utilizing pore pressure prediction from different logging/measurement parameters. For the other methodology which utilizes resistivity and porosity for pore pressure prediction, the methodology is proposed to utilize the change in Archie’s cementation exponent. This is because the effect of cementation on pore pressure prediction could become significant at greater depths. Testing with field data showed that this methodology also performs better than simply averaging pore pressure prediction from resistivity and porosity logs using other conventional equations. In addition to the methodology developed for combining porosity and resistivity log for pore pressure prediction, machine learning can also be utilized. Results obtained using artificial neural network indicate better performance in comparison to simply averaging predictions from conventional means of using resistivity and porosity.