Synthetic Gas-Oil-Ratio Determination Using Genetic Algorithms
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional predictive modeling of Gas-Oil-Ratios (GORs) faces challenges in determining GORs for gas dominant fluids, as actual values are often unknown or hard to measure, especially for pure gas GORs which are theoretically infinity, making calibration and prediction difficult.
Innovation Solution
The approach involves using a genetic algorithm and multivariate regression simulator to iteratively assign synthetic GOR values, combining fluid compositional concentrations and optical sensor responses, and applying a neural network algorithm to improve GOR model predictions across a database of global fluid samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional predictive modeling is used for GOR determination, then GOR can be predicted for typical fluids, but accurate determination for gas dominant fluids fails because actual values are unknown or hard to measure
Solution Approach 1:
The patent creates a synthetic GOR database that copies the structure and relationships of real GOR data, using simulated optical responses and compositional concentrations to generate realistic GOR values for gas dominant fluids. This synthetic copy allows the model to learn from fabricated data that mimics actual measurement conditions without requiring difficult field measurements.
Solution Approach 2:
The patent transforms the GOR determination problem by changing the parameter space to include compositional concentrations and optical sensor responses as intermediate parameters. By optimizing these parameters through genetic algorithms and neural networks, the system indirectly determines GOR values without directly measuring them, thus avoiding the measurement difficulty.
2Reliability
If synthetic GOR values are assigned to gas dominant fluids, then model calibration improves, but the complexity of the determination process increases due to iterative optimization requirements
Solution Approach 1:
The patent implements a self-calibrating system where the neural network automatically adjusts its parameters using the synthetic GOR database. The genetic algorithm performs self-optimization by iteratively improving the synthetic GOR assignments based on model performance feedback, eliminating the need for manual calibration and reducing operational complexity despite the sophisticated algorithms used.
Solution Approach 2:
The system incorporates feedback loops where the neural network's predictions are compared against the synthetic GOR values, and the genetic algorithm uses this feedback to refine the synthetic database. This continuous feedback mechanism improves model reliability iteratively while automating the complexity management through algorithmic control.
Data Source
AI summary
The disclosed embodiments include a method, apparatus, and computer program product for determining a synthetic gas-oil-ratio for a gas dominant fluid. For example, one disclosed embodiment includes a system that includes at least one processor, and at least one memory coupled to the at least one processor and storing instructions that when executed by the at least one processor performs operations that include optimizing a gas-oil-ratio database using a genetic algorithm and a multivariate regression simulator and generating a synthetic gas-oil-ratio for a gas dominant fluid. In one embodiment, optimizing a gas-oil-ratio database using a genetic algorithm and a multivariate regression simulator comprises defining gas-oil-ratio searching boundaries gas-oil-ratio for each gas dominant fluid; assigning randomly a synthetic gas-oil-ratio for each gas dominant fluid in a set of gas dominant fluids in the initial population of gas-oil-ratio data, wherein the gas-oil-ratio for each gas dominant fluid is within the searching boundaries; generating an initial population of gas-oil-ratio data for a set of gas dominant fluids; and evaluating synthetic gas-oil-ratio assignments for the initial population using the multivariate regression simulator.