Virtual Gas Metering Using Machine Learning for Accurate Well Flow Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating gas flow rates in oil and gas fields are inaccurate due to simplifying assumptions and lack of sufficient sensory inputs, leading to unreliable predictions and high costs associated with physical meters.
Innovation Solution
A machine-learned model, specifically a support vector regression method, is employed to estimate gas flow rates using real-time sensor data from field instruments, allowing for continuous updates and accurate predictions without the need for physical meters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physical gas meters are used to measure gas flow rate, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent creates a virtual copy of the physical gas meter by training a machine learning model on data from physical meters and field instruments. The trained model reproduces the measurement function without requiring the actual physical meter hardware, thereby eliminating the need for expensive physical meters while maintaining measurement capability.
Solution Approach 2:
The patent replaces the mechanical/physical gas metering system with a computational model that uses machine learning algorithms. The system substitutes physical measurement hardware with a software-based prediction model that processes data from field instruments to estimate gas flow rates, thereby reducing device complexity and cost.
2Measurement precision
If physical gas meters are installed at every well, then measurement precision is improved, but loss of substance (cost) increases
Solution Approach 1:
The patent creates a virtual copy of the physical gas meter by training a machine learning model on data from physical meters and field instruments. The trained model reproduces the measurement function without requiring the actual physical meter hardware, thereby eliminating the need for expensive physical meters while maintaining measurement capability.
Solution Approach 2:
The patent replaces expensive physical gas meters with a computational model that can be updated and refined. The model serves as a cost-effective alternative that eliminates the need for costly hardware infrastructure while providing continuous measurement capability through software updates rather than physical replacements.
3Device complexity
If simplified methods are used to estimate gas flow rate, then device complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The patent implements a feedback mechanism where the machine learning model continuously learns from new data and improves its predictions. The model is trained on historical data from physical meters and field instruments, and can be retrained or fine-tuned as new data becomes available, thereby maintaining high measurement precision while keeping the system relatively simple.
Solution Approach 2:
The patent changes the approach from using simple physical measurement equations to using machine learning models that process multiple parameters from field instruments. The model considers various input parameters such as pressure, temperature, and flow conditions to make accurate predictions, thereby improving measurement precision without significantly increasing device complexity.
Data Source
AI summary
A method including receiving modeling data, wherein the modeling data includes field instrument data and associated gas flow rate data for a first plurality of wells. The method further includes splitting the modeling data into a train set, a validation set, and a test set and pre-processing the modeling data. The method further includes selecting a machine-learned model and architecture and training the machine-learned model to predict gas flow rates from the pre-processed modeling data using the training set. The method further includes using the machine-learned model to predict gas flow rates using real-time field instrument data from a second plurality of wells, wherein the real-time field instrument data has been pre-processed similarly to the modeling data.