Machine Learning Application on Formation Testing Result Status Prediction – A Case Study

Abstrak

Data-driven recommendations and decisions play an important role in today’s challenging oil and gas industries. Incorporating automation offers the ability to analyze bigger, more complex data and deliver fast results – even on a very large scale. Machine learning is one of the branches of Artificial Intelligence which can automate the analytical model building by learning from data, identify patterns and make decisions with minimal human intervention. On this study, machine learning method was applied on a set of data from one field to predict the pressure test or pretest status of wireline formation testing result for a given depth with a model built based on the well logs. First, rock quality classification was performed using an unsupervised approach. Next, using the previous step output as a constraint, Neural Networks method was applied to predict the status of wireline formation testing pretest result. The result of pretest prediction was then validated with actual pretest result. This study is using the dataset of four wells as “train” data and predicting the result status of one “target” well. The accuracy of prediction of the “train’ dataset was above 80% and the accuracy for the “target” well was 100%. This difference might be explained by the wide range of data that was incorporated as “train” dataset wells which able to build a robust model to predict the “target” well accurately. This study shows the application of machine learning on good data set will leverage the value of data providing different perspective from the conventional decision-making process.