ESP Wells Dynamic Survival Analysis and Lifespan Prediction Using Machine Learning Algorithms

Abstract The electric submersible pump (ESP) is the primary artificial lift method for offshore oil wells. Due to the high operation and maintenance costs, it is crucial to avoid operations and conditions that negatively impact ESP wells' lifespan and seek ways to prolong their run life. However, the lifespan of ESP is affected by many factors, and it isn't easy to quantify the effects of these factors by conventional methods. To improve the lifespan of ESP and identify critical factors affecting it, this study thoroughly utilizes the wealth of downhole and wellhead measurement parameters available for ESP wells, establishes a dynamic survival model based on Cox proportional hazards survival analysis, and further constructs a Long Short-Term Memory (LSTM) network that can predict the remaining life based on survival modeling. Field application results indicate that the LSTM network can accurately predict the remaining life of ESP and that the factors influencing ESP lifespan can be identified and quantified by the survival analysis model.