ESP Back Pressure Valve Control Using Machine Learning Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing ESP systems face challenges with power consumption, control precision, and adaptability to changing downhole conditions, leading to increased operational costs and suboptimal hydrocarbon recovery rates, with traditional optimization methods requiring manual intervention and lacking full autonomy.
Innovation Solution
A back pressure control valve system with a programmable logic controller and machine learning model that continuously adjusts actuator settings to minimize error around a set point, using telemetry data from various sensors to optimize fluid flow and pump frequency, enabling semi-autonomous or autonomous operation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional PID control is used for back pressure control, then the system is simple to implement, but the control precision and adaptability to changing downhole conditions are insufficient
Solution Approach 1:
The system implements continuous feedback loops where telemetry data from downhole sensors is constantly monitored, processed by machine learning models, and used to adjust back pressure control valve settings in real-time. This closed-loop feedback mechanism enables precise adaptation to changing downhole conditions while maintaining system stability.
Solution Approach 2:
The machine learning models autonomously analyze telemetry data, predict optimal control parameters, and adjust valve settings without requiring manual intervention. The system self-optimizes by learning from historical data and adapting to new conditions automatically, reducing the need for human expertise while improving control precision.
2Productivity
If manual ESP optimization is performed, then expert knowledge can be applied, but the process is time-consuming and cannot achieve continuous optimization
Solution Approach 1:
The system enables continuous optimization by constantly monitoring telemetry data, processing it through machine learning models, and adjusting control parameters without interruption. Unlike manual optimization which occurs periodically, the automated system operates continuously, ensuring optimal performance at all times and maximizing hydrocarbon recovery.
Solution Approach 2:
The patent replaces manual expert analysis and adjustment with automated machine learning algorithms. The ML models process telemetry data and generate control commands automatically, substituting the mechanical process of human intervention with an automated computational system that operates faster and without fatigue.
3Reliability
If frequent valve actuator adjustment is implemented, then tight control around set point is achieved, but the system complexity and computational requirements increase
Solution Approach 1:
The system dynamically adjusts the frequency and magnitude of valve actuator corrections based on real-time conditions. The machine learning models continuously evaluate telemetry data and modify control commands accordingly, enabling tight control around the set point while adapting to changing downhole conditions without requiring overly complex predetermined control schedules.
Data Source
AI summary
On an oil well equipped with an electric submersible pump (ESP), a valve combination coupled with a controller running a machine learning model provides for continuous valve actuator setting adjustment of a back pressure control valve (e.g., valve actuator settings updated at intervals on the order of 100 ms), and/or adjustment of a variable speed drive (VSD), to achieve tight control around a tubing pressure set point for the well. The system for ESP control may be network-connected, e.g. to the Internet, and pump and well data and the external tubing pressure set point may be communicated through the network.