Beicip-Franlab will present the following talk at 5th AAPG Siliciclastic Reservoir of the Middle East GTW.
The field of petrophysical interpretation plays an important role in the exploration and production of hydrocarbons. Accurate assessment of reservoir properties such as porosity, permeability, and fluid saturation are essential for optimizing drilling, production, and reservoir management strategies.
Traditionally, petrophysicists have relied on well-established models and manual interpretation methods to extract valued information from well logs and geological data. The keys steps in conducting a petrophysical interpretation study involve log database creation, log construction, petrophysical interpretation, electrofacies and rock-typing.
Artificial Intelligence (AI) and Machine Learning (ML) have emerged as transformative technologies offering the potential to accelerate and revolutionize petrophysical interpretation study.
In this paper, we would like to introduce a transformative digital workflow which employ Artificial Intelligence (AI) and Machine Learning (ML) at every stage of petrophysical interpretation task. This workflow not only enhances accuracy and efficiency but also incorporates an operational aspect to facilitate its practical utilization.
The first step in petrophysical interpretation study is to establish a robust log database. This task involves collecting and organizing various types of well log data together with core measurement. Principal logs used for petrophysical interpretation (GR, NPHI, RHOB, DT, RES….) are selected. The collected data are qualitychecked for accuracy, depth matching, and consistency. This can be a particularly time-consuming process,especially when dealing with thousands wells and scattered log data across various folders. Artificial intelligence and Machine Learning can be particularly useful in this phase. Thousands of las/dlis file can be automatically analyzed by using open-source Python libraries. With use of Natural Languages Processing, log data can be classified into three distinct categories: Cased hole log (CHL), logging while drilling (LWD) and Wireline log (WL). The selection of LWD and WL logs for petrophysical interpretation is facilitated by rule-based expert systems, while outlier detection machine learning algorithms can help to detect log quality issues. The process also automates the creation of a log availability summary, making it convenient to review the data.
