Real Time Petrophysics via Artificial Intelligence in Brown Field Development

Abstract

The paper introduces a novel approach to evaluate well log data in real time during downhole data acquisition. This approach has been successfully implemented in multiple drilling projects. It relies on effective real-time data transmission with the key enabler of the system is development of the well log artificial intelligence module. Time is of the essence during drilling operations. Hence, any efforts towards increasing the operation efficiency can lead to a safer and more optimized well delivery. A machine learning application for real-time petrophysical evaluation during drilling was successfully developed to achieve this ambition. The application which named Well Log Data Artificial Intelligence was tested in a Malaysia complex brownfield infill drilling project, where it proved to promote a seamless decision-making process during drilling and simultaneously lead to financial and operation value creation. The application uses selected key wells in the field as training dataset before the actual drilling operation. Random Forest ensemble was utilized for model learning and correspondent petrophysical evaluation, The module can predict reservoir characteristics in real-time by leveraging on feeder open hole raw curves as inputs with integration of existing trained model and real time data transmission. Moreover, the application provides easily accessible prediction results through the same display interface as the real-time data transmission. This makes it attractive for multi-disciplinary utilization during discussion.