AI-Based Approach for ESP Optimization

Abstract

Abstract ESP performance and health optimization play a crucial role in maximizing oil well production and overall profitability. The selection of appropriate control set points, such as pump frequency and tubing pressure, and the intervals at which adjustments are made, significantly impact production outcomes. This paper introduces an AI-based approach for ESP optimization that operates in near real-time, allowing for dynamic set point adjustments based on varying well conditions. A dataset comprising telemetry, production, and statistical features from 193 ESP-operated wells was utilized to train the optimization model. The dataset covers a total duration of 67,358 days, averaging approximately 349 days per well. The telemetry feature set encompasses 23 distinct values, while additional information, such as well age, peak production rates, theoretical production rates, and other pertinent factors, was incorporated to provide valuable insights into various well characteristics. By leveraging this extensive dataset and employing AI techniques, the proposed approach enables effective ESP optimization, improving production outcomes and overall well profitability.