Abstract:A virtual flow meter (VFM) enables continuous prediction of flow rates in petroleum production systems. The predicted flow rates may aid the daily control and optimization of a petroleum asset. Gray-box modeling is an approach that combines mechanistic and data-driven modeling. The objective is to create a computationally feasible VFM for use in real-time applications, with high prediction accuracy and scientifically consistent behavior. This article investigates five different gray-box model types in an industrial case study using real, historical production data from 10 petroleum wells, spanning at most four years of production. The results are diverse with an oil flow rate prediction error in the range of 1.8%-40.6%. Further, the study casts light upon the nontrivial task of balancing learning from both physics and data. Consequently, providing general recommendations towards the suitability of different hybrid models is challenging. Nevertheless, the results are promising and indicate that gray-box VFMs may reduce the prediction error of a mechanistic VFM while remaining scientifically consistent. The findings motivate further experimentation with gray-box VFM models and suggest several future research directions to improve upon the performance and scientific consistency.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: How to create a computationally feasible model with high prediction accuracy and scientifically consistent behavior for the virtual flow meter (VFM) in the oil production system through the grey - box modeling method. Specifically, the research aims to balance the physics - based and data - driven modeling methods to optimize the daily control and optimization of oil assets, especially in terms of flow rate prediction of multiphase flow (gas, oil, water mixtures). ### Research Background In the oil production system, the virtual flow meter (VFM) is used to continuously predict the flow rate and help optimize the daily control of oil assets. However, the characteristics of multiphase flow make flow rate prediction complex. Traditional VFM models are divided into two categories: 1. **Physics - based white - box models**: These models rely on prior knowledge of the process and first - principle laws. Although they have high scientific consistency, they may lack flexibility in real - time applications. 2. **Data - driven black - box models**: These models use patterns in historical data for prediction. Although they are highly flexible, they lack scientific consistency and are sensitive to data quality and quantity. ### Research Objectives To overcome the limitations of the above two models, this paper proposes the grey - box modeling method, which combines the advantages of physics - based and data - driven methods. Specific objectives include: - **Improve prediction accuracy**: By introducing data - driven elements, make up for the errors caused by simplifications and assumptions in physics - based models. - **Maintain scientific consistency**: Ensure that the model output conforms to known scientific principles. - **Achieve real - time application**: Ensure that the model is computationally efficient enough for real - time control and optimization. ### Research Methods This paper achieves its objectives through the following steps: 1. **Select five different grey - box model types**: Each model combines physics - based and data - driven methods to different degrees. 2. **Use real historical production data**: Data from 10 oil wells, covering production records of up to four years. 3. **Evaluate model performance**: Evaluate the performance of different models in practical applications by comparing their prediction errors. ### Main Findings The research shows that the grey - box VFM model can significantly reduce prediction errors while maintaining scientific consistency. Specifically, the oil flow rate prediction error ranges from 1.8% - 40.6%, indicating that the grey - box model can be superior to pure physics - based models in some cases. However, balancing the relationship between physics - based and data - driven methods remains a challenge, so it is difficult to give general recommendations. ### Future Research Directions This research provides a basis for further exploration of the grey - box VFM model and points out several future research directions, including but not limited to: - **Improve model structure**: Develop more complex grey - box models to adapt to more types of oil production systems. - **Optimize parameter estimation**: Improve parameter estimation methods to better balance the influence of physics - based and data - driven methods. - **Expand application scenarios**: Apply the grey - box model to other types of industrial systems to verify its generalization ability. In conclusion, through in - depth research on the grey - box modeling method, this paper provides a new solution for flow rate prediction in the oil production system and shows its potential in improving prediction accuracy while maintaining scientific consistency.

On gray-box modeling for virtual flow metering

When is gray-box modeling advantageous for virtual flow metering?

Developing a Hybrid Data-Driven, Mechanistic Virtual Flow Meter -- a Case Study

The VOF Method Based on Refined Grids Partition of Partial Domain

Virtual Multiphase Flowmetering Using Adaptive Neuro-Fuzzy Inference System (ANFIS): A Case Study of Hai Thach-Moc Tinh Field, Offshore Vietnam

Grey-box models for wave loading prediction

Fault Prediction Using a Grey-Markov Model from the Dissolved Gases Contents in Transformer Oils

Virtual Meter with Flow Pattern Recognition Using Deep Learning Neural Networks: Experiments and Analyses

Real-Time Liquid Rate and Water Cut Prediction From the Electrical Submersible Pump Sensors Data Using Machine-Learning Algorithms

Bayesian Nonlinear Gaussian Mixture Regression and Its Application to Virtual Sensing for Multimode Industrial Processes.

The Gray-Box Based Modeling Approach Integrating Both Mechanism-Model And Data-Model: The Case Of Atmospheric Contaminant Dispersion

Development of Machine Learning-Based Production Forecasting for Offshore Gas Fields Using a Dynamic Material Balance Equation

Physics Constrained High-Precision Data-Driven Modeling for Multi-Path Ultrasonic Flow Meter in Natural Gas Measurement

A Machine Learning Approach for Virtual Flow Metering and Forecasting

Flow Fusion, Exploiting Measurement Redundancy for Smarter Allocation

A Heterogeneous Model for Simulating Fluid Flow in Naturally Fractured-Vuggy Carbonate Reservoirs

The Benefit of Noise-Injection for Dynamic Gray-Box Model Creation

Multiphase friction calculation peculiarities based on models implemented in the "d-Flow" software package

Stochastic Modelling of the Flow-Front Evolution in a Vacuum Assisted Resin Transfer Moulding Process with Missing Data

Virtualized Viscosity Sensor for Onboard Energy Management

An Efficient Hybrid Model for 3D Complex Fractured Vuggy Reservoir Simulation