By continuing to browse this website, you agree to our use of cookies. Learn more at the Privacy Policy page.
Contact Us
Contact Us
ML-based virtual flow meter
  • High load
  • AI & ML

ML-based virtual flow meter

Download the case study
Custom enterprise-grade AI and data solutions for the Oil & Gas sector

Client background

The client is an American multinational oilfield services company with operations in 75+ countries, providing innovative solutions for oil and gas production, drilling, and reservoir management.

Before implementing the virtual metering system, the client faced several operational challenges:

  • Reliance on expensive MPFMs that incurred high installation and maintenance costs.
  • Frequent failures and calibration issues, leading to production delays.
  • Limited scalability due to high costs per physical meter.
  • Inconsistent data from different well conditions, making accurate flow estimations difficult.

The company needed an ML-driven virtual metering system to replace or supplement physical flow meters, reduce costs, and improve efficiency.

Business challenge

The company sought an AI-driven virtual metering solution to improve efficiency, cost savings, and operational insight across its pipeline infrastructure.

Potential threat:
If the solution was not implemented, the client would continue to face rising operational costs, unplanned downtime, and inefficiencies in production monitoring.

Constraints:

  • Physical: Diverse well conditions affecting flow measurement consistency.
  • Infrastructure: Compatibility with existing SCADA and IoT systems.
  • Financial: Budget constraints required a scalable and cost-effective solution.
  • Data: Lack of labeled training data for ML models.
  • Technical: Sensor noise and anomalies leading to inaccurate readings.
  • Regulatory: Compliance with industry safety and environmental standards.
  • Technological: Ensuring real-time processing and model explainability.

Problems and solutions

The virtual flow metering system presented a number of engineering and data science challenges typical of industrial-grade ML deployments. Xenoss team had to create bespoke solutions to overcome infrastructure, data, and performance limitations.

Heterogeneous sensor data

The solution had to process a wide variety of sensor types (temperature, pressure, vibration, acoustic), each with unique formats and noise characteristics.

Solution: We implemented a flexible data ingestion and preprocessing pipeline capable of normalizing and validating diverse input formats in real time. The system supports secure, scalable streaming from edge devices and sensors, ensuring consistent data quality across all environments.

Limited labeled data for ML training

The lack of labeled flow rate data made it difficult to train supervised models from scratch.

Solution: We applied a hybrid modeling approach, combining physics-based models with ML to reduce reliance on labeled data and maintain consistency with physical constraints.

Dynamic and transient flow regimes

Multiphase pipelines often exhibit non-steady behavior, such as slug flow or abrupt regime changes, which are difficult for static models to capture.

Solution: Our pipeline incorporated LSTM-based time-series models alongside first-principles equations to handle transitions and produce stable predictions under transient conditions.

Real-time inference and SCADA integration

Ensuring sub-second inference latency and compatibility with existing SCADA systems required architectural optimization.

Solution: We deployed models within a cloud-native inference framework, using event-triggered execution and lightweight serverless functions to ensure low-latency predictions. The system exposed secure API endpoints for seamless integration with external platforms, enabling real-time responsiveness across diverse operational environments.

Sensor drift and inconsistent data quality

Over time, physical sensors (especially in harsh oilfield environments) degrade, introduce noise, or drift from calibration, resulting in reduced model accuracy.

Solution: We implemented automated data validation, anomaly detection, and reconciliation rules based on conservation laws (mass/energy balance). This enabled ongoing correction of flawed sensor inputs without human intervention.

Model drift and lifecycle management

ML models degrade as operating conditions, fluid properties, or equipment configurations change, making predictions stale.

Solution: We built an MLOps pipeline to monitor model performance, detect drift, and automatically trigger retraining cycles using fresh data. This ensured sustained model accuracy across changing operational regimes.

Architectural highlights

Scalable hybrid architecture

The virtual flow meter system integrated physics-based and ML models in a horizontally scalable design, enabling rapid deployment across new wells with minimal customization.

Real-time inference & anomaly detection

Real-time predictions were delivered through an event-driven inference pipeline with built-in anomaly detection, generating instant flow estimates from incoming sensor data to support timely decisions.

Sensor fusion & efficient data streaming

Multiple sensor streams (pressure, temperature, acoustic, vibration) were ingested through a scalable streaming data layer with real-time signal enrichment, enabling deeper insights without added latency.

Robust monitoring & retraining pipeline

A fully automated MLOps workflow ensured continuous model monitoring and retraining on data drift, while real-time dashboards gave operators full visibility into predictions, system health, and anomalies.

Hybrid modeling engine

At the core is a hybrid engine combining first-principles fluid dynamics with machine learning, delivering accurate flow estimates across steady and dynamic conditions with minimal recalibration.

Secure, SCADA-integrated deployment

The system seamlessly integrates with industrial SCADA and IoT environments using standard protocols, while built-in security features, including role-based access, encryption, and network isolation, ensure compliance with safety and data protection standards.

Technical and MLOps approach

Data validation and reconciliation to ensure accurate flow measurement by aligning sensor inputs with physical constraints like mass conservation and thermodynamics.

Automated screening & correction

Statistical checks detect outliers or missing values.

Constraint-based optimization

Tuning model parameters (e.g., friction factors, choke discharge
coefficients) to match established operating envelopes.

Comparison with physical meters

Cross-check predictions against multiphase flow meters or test
separators for periodic calibration where available.

Multiphase flow modeling to capture the complex interactions of oil, gas, and water flowing together in dynamic pipeline conditions.

Core considerations

  • Oil, gas, and water phase interactions vary greatly by GOR, water cut, temperature, and pressure.
  • Special caution is necessary in fields prone to slugging or abrupt flow changes.

Hybrid deployment

Blending mechanistic flow correlations with ML to handle real-time correction factors.

Adaptive tuning

Retune model parameters automatically if fluid compositions or operating regimes shift over time.

Soft sensing to virtually estimate parameters that are costly or impractical to measure directly, turning raw signals into actionable insights.

Techniques

  • Neural Networks (CNN/MLP) for direct sensor-based flow prediction.
  • Recurrent architectures (LSTM/RNN) for dynamic or transient phenomena.
  • Ensemble or meta-learning methods to consolidate multiple model outputs.

Model lifecycle (MLOps)

  • Automated retraining triggered by new data or drift detection.
  • Version control and rollback for consistent reliability in production.

Cloud-based deployment enables real-time operations, scalable data processing, and automated model retraining, allowing for efficient and secure adaptation to changing conditions.

Scalability

Expanding to additional wells and fields with minimal overhead.

Security & compliance

Employing Identity and Access Management, encryption, and network isolation for critical infrastructure data.

Continuous integration/delivery

Automated retraining and redeployment pipelines ensure the system remains up to date and responsive to changing data and operating conditions.

Results

40% OpEx reduction

Less reliance on costly multiphase flow meters and less frequent test separator usage.

95%+ accuracy in flow estimation

Constraint-based optimization ensures that final flow predictions reflect established fluid and operational physics.

Zero dependency

on expensive MPFMs in select sites.

Scalability

A cloud-native MLOps pipeline for deploying VFMs across multiple fields.

Real-time insights

Automated dashboards and anomaly alerts that help operators intervene proactively reduced data inconsistencies by 60%.

30% downtime reduction

Enabled by early anomaly detection and real-time operational insights.

Want to build your own solution?

Contact us