Autonomous Well Performance Troubleshooting; A Promising Data-Driven Application

Abstract

Well Performance can deteriorate due to several reasons, for example: formation damage, scale buildup, back pressure from other wells, artificial lift issue etc. In this paper we present an application of utilizing machine learning to build a model to articulate and flag deterioration and reason behind it. The model was used to flag problems such as salt and scale build up in the tubing as well as backpressure due to emulsions in the tubing or in topside pipes. The model was capable of monitoring well performance using only the well head parameters