Analysis of candidate oil wells for workover interventions using machine learning tools

Throughout their production life oil wells have the potential to undergo workover interventions with one of the possible objectives to maintain healthy production levels. In mature fields, generally in a declining production stage, increases the need of evaluations and monitoring of the flow levels of productive fluids, to indicate those who are producing below their potential. In this context, a study case was conducted to analyze operational parameters associated with well production fluids by applying data science tools, considering the premise all are intact in terms of safety barriers. Starting from the selection of interest variables and data collection, an exploratory analysis was initially proposed to examine the interrelations between variables. For the identification of patterns and well groupings, two machine learning techniques were applied: Principal Component Analysis (PCA) and Cluster Analysis. PCA allowed the reduction of the variable set dimensionality, facilitating a practical visualization of each well’s productive scenario based on the analysis of the three selected principal components. Furthermore, through Cluster Analysis, wells were categorized into five distinct Clusters, enhancing the monitoring and management of potential candidate wells.