Model-Constrained Multi-Phase Virtual Flow Metering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for forecasting multi-phase flow rates in oil and gas production are inaccurate and inflexible, failing to account for complex geological and reservoir data, dynamic operation management events, and rapid production rate changes, often requiring extensive domain knowledge and expensive physical experiments.
Innovation Solution
A computer-implemented method using machine learning to build model-constrained multi-phase virtual flow metering and forecasting, combining unconstrained flow models, well dynamics models, and virtual sensing models to forecast future target flow rates and predict real-time multi-phase flow rates, incorporating historical and auxiliary sensor data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional methods are used for forecasting multi-phase flow rates, then the process is simpler, but the accuracy and reliability of forecasting are poor
Solution Approach 1:
The forecasting system is segmented into multiple specialized models: unconstrained flow models for future target flow rates, well dynamics models for auxiliary sensor data, and virtual sensing models for real-time flow rates. Each model handles specific aspects of the forecasting problem, improving overall accuracy while managing complexity through modular architecture.
Solution Approach 2:
The patent combines multiple modeling approaches (unconstrained flow models, well dynamics models, and virtual sensing models) into a unified constrained forecasting model. This integration leverages the strengths of each component model to achieve superior forecasting accuracy that cannot be obtained by any single model alone.
2Adaptability or versatility
If traditional forecasting methods are used, then the system is easier to operate, but the adaptability to complex geological and reservoir data is insufficient
Solution Approach 1:
The forecasting system employs dynamic models that can adapt to changing conditions. The well dynamics models capture time-varying behavior of the well system, while the virtual sensing models are trained on historical data to learn complex non-linear relationships. This dynamic approach enables the system to adapt to complex geological and reservoir data while maintaining automated operation.
Solution Approach 2:
The system utilizes multiple parameters including current and historic multi-phase flow rates, auxiliary sensor data, and predicted target flow rates as inputs to the models. By considering changes in multiple parameters simultaneously, the system achieves high adaptability to complex data while the automated model construction reduces operational complexity.
3Reliability
If traditional methods are used, then the implementation is faster, but the reliability and accuracy of production forecasts are insufficient
Solution Approach 1:
The system performs preliminary actions by training unconstrained flow models, well dynamics models, and virtual sensing models on historical data before actual forecasting is needed. This pre-training establishes a reliable foundation that can be quickly applied to new forecasting scenarios, improving reliability while reducing the time needed for actual production forecasts.
Solution Approach 2:
The virtual sensing models create virtual copies of physical flow measurements by learning from historical data and sensor inputs. These virtual flow rate predictions replicate the behavior of physical flow meters without requiring physical experimentation, thereby improving forecast reliability while reducing the time and resources needed for physical validation.
4Measurement precision
If unconstrained flow models are used alone, then the forecasting is faster, but the physical consistency and accuracy of predictions are poor
Solution Approach 1:
The virtual sensing models serve as intermediaries that translate predictions from unconstrained flow models and auxiliary predictions from well dynamics models into physically consistent flow rate predictions. These intermediary models ensure that the final predictions satisfy physical constraints while maintaining the computational efficiency of the unconstrained models.
Solution Approach 2:
The system replaces direct physical experimentation and complex mechanical measurement systems with machine learning models. The virtual sensing models substitute for physical flow meters by using data-driven approaches to predict flow rates, thereby improving accuracy while managing the complexity of model integration through automated training procedures.
Data Source
AI summary
A computer-implemented method for constrained multi-phase virtual flow metering and forecasting is described. The method includes predicting instantaneous flow rates and forecasting future target flow rates and well dynamics. The method includes constructing a virtual sensing model trained using forecasted target flow rates and well dynamics. The method includes building a constrained forecasting model by combining unconstrained flow forecasting models, well dynamics models, and virtual sensing models, wherein the constrained forecasting model forecasts multi-phase flow rates.