A Machine Learning Approach Related to Well-Log Correlation

Machine Learning has high potential and can be applied to predict and optimize various processes. It has many applications for complex tasks and those that people do every day. Machine Learning is a subfield of Artificial Intelligence. Machine Learning can be used to find and analyze data, recognize patterns, and extract materials from internet. There are many forms of Machine Learning; supervised, unsupervised, semi-supervised, and reinforcement. The model is composed of a dataset and a model is set up to train the known data. Then the model trains and learns more in order to predict unknown outcomes. It is consistently tuned to reevaluate known outcomes to make predictions as accurate as possible.In the petroleum industry it can be applied in many forms such as geological studies, maintenance of petroleum equipment, and production engineering. This research is developed around modeling well-log data to optimize and make predictions for the future using stratigraphic correlation techniques. Common type of well-log data is resistivity versus depth. Normally, the higher the resistivity the less water saturation. Water at deeper depths contains a lot of salt so it is highly conductive, which means less hydrocarbons are present. However, higher resistivity generally correlates to more hydrocarbons being present. We are currently working on the model that should recognize the well-log patterns. It is intended to (1) be used to interpret well data, (2) being trained on added datasets, (3) make predictions for oil and gas production optimization from new or existing wells.