Multiphase Flow Rate Cleansing for Reliable VFM Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing Virtual Flow Metering (VFM) models rely heavily on high-quality ground truth labels for training, which are often compromised by sensor malfunctions, missing data, or incorrect readings, making it difficult to achieve accurate multiphase flow rate measurements in oil and gas production.
Innovation Solution
A method and system utilizing multiple machine-learned models to preprocess and cleanse multiphase flow rate data from sensors, incorporating auxiliary data to generate high-quality flow rate data for training VFM models, ensuring accurate flow rate predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If real-world sensor data is used for VFM model training, then data authenticity and practical applicability are improved, but data quality deteriorates due to sensor malfunctions, missing readings, and incorrect measurements
Solution Approach 1:
The patent introduces an intermediary data cleansing system that mediates between raw sensor data and VFM model training. This system uses multiple machine-learned models (first model for initial cleansing, second model for auxiliary data integration, third model for final high-quality data generation) to process and clean real-world sensor data, transforming unreliable raw data into reliable training data while preserving its practical applicability
Solution Approach 2:
The patent applies parameter changes by transforming the state of raw data through multiple processing stages. The first model changes parameters by identifying and correcting erroneous readings, the second model changes parameters by integrating auxiliary sensor data, and the third model changes parameters by generating high-quality cleansed data suitable for VFM training, thereby improving data quality while maintaining authenticity
2Device complexity
If traditional single-model data cleansing is used, then system simplicity is maintained, but data cleansing effectiveness deteriorates due to inability to handle complex sensor errors and missing data
Solution Approach 1:
The patent segments the data cleansing process into three distinct models with specialized functions: the first model handles initial data cleansing and error detection, the second model processes auxiliary sensor data and integrates it with primary measurements, and the third model generates final high-quality cleansed data. This segmentation allows each model to specialize in specific cleansing tasks, dramatically improving data cleansing effectiveness while maintaining manageable system complexity through modular architecture
Solution Approach 2:
The multi-model system achieves universality by creating a comprehensive data cleansing framework that can handle various types of sensor errors, missing data, and anomalies through different specialized models. Each model serves multiple functions within its domain, and collectively they provide universal data cleansing capability across diverse sensor failure modes and data quality issues
Data Source
AI summary
A method includes obtaining raw multiphase flow rate data from a first set of sensors disposed on a pipeline of a well and obtaining auxiliary data from a second set of sensors disposed on the pipeline. The method further includes preprocessing the raw multiphase flow rate data to form a preprocessed multiphase flow rate dataset (“preprocessed dataset”), and determining, with a first model processing the preprocessed dataset, a first cleansed flow rate dataset. The method further includes determining, with a second model processing the preprocessed dataset and the auxiliary data, a second cleansed flow rate dataset. The method further includes determining high quality flow rate data with a third model processing the preprocessed dataset, the first cleansed flow rate dataset, and the second cleansed flow rate dataset. The method further includes transmitting the high quality flow rate data to a control system of the well.