Produced Water Origin Classification Using Elemental Ratios and Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining the origin of produced water in oil and gas fields are unreliable and inaccurate, as they rely on chloride and salinity concentrations, which can overlap between natural and artificial sources, and are costly, failing to differentiate between mixed content or inflow from multiple sources.
Innovation Solution
A method involving the collection and analysis of geochemical data from produced water samples, combined into a database, and processed using a machine-learned model to classify the water type, utilizing elemental ratios and derived features to accurately predict the origin of produced water samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If chloride and salinity concentrations are used to determine the origin of produced water, then the method is simple to implement, but the accuracy and reliability of origin determination deteriorates due to overlapping concentrations between natural and artificial sources
Solution Approach 1:
The patent transitions from using single parameters (chloride and salinity concentrations) to using multiple parameters including elemental ratios (e.g., Br/Ca, Sr/Ba, Na/K) and derived features. This multi-parameter approach enables more accurate discrimination between natural and artificial water sources by capturing the complex geochemical fingerprints of different water types.
Solution Approach 2:
The patent replaces traditional manual geochemical analysis methods with machine learning algorithms that automatically process geochemical data and classify water origins. The system uses trained models to predict water type probabilities, substituting expert interpretation with automated computational analysis that improves consistency and accuracy.
2Quantity of substance
If traditional geochemical analysis methods are used, then the cost is reduced, but the ability to differentiate between mixed content or inflow from multiple sources deteriorates
Solution Approach 1:
The patent adds dimensional complexity to the analysis by incorporating multiple elemental ratios and derived features beyond simple concentration measurements. This multi-dimensional approach creates a more robust geochemical fingerprint that can distinguish mixed water sources and multiple inflow origins that single-parameter methods cannot resolve.
Solution Approach 2:
The patent introduces machine learning models as intermediaries that process and interpret complex geochemical data relationships. These models act as mediators between raw geochemical measurements and origin determination, uncovering subtle patterns and interactions in the data that reveal mixed content and multiple sources without requiring additional physical measurements.
3Measurement precision
If machine learning models are used to classify produced water types, then the accuracy and reliability of origin determination improves, but the complexity of the system increases
Solution Approach 1:
The patent extracts the complex decision-making logic from the overall system by encapsulating it within trained machine learning models. Once trained, these models function as self-contained black boxes that require minimal human intervention during operation, effectively removing the complexity of manual interpretation while maintaining high accuracy in origin determination.
Solution Approach 2:
The patent performs preliminary action by training the machine learning models offline using comprehensive training datasets before deployment. This pre-training phase captures complex relationships in the data, allowing the models to make accurate predictions during operation without requiring complex real-time processing or human expertise, thereby reducing operational complexity while maintaining high accuracy.
Data Source
AI summary
A method involving collecting a first geochemical data set for a first plurality of produced water samples; collecting a second plurality of produced water samples; performing geochemical analyses on the second plurality of produced water samples to form a second geochemical data set; and combining the first and second geochemical data sets into a database. The method further includes determining, by a subject matter expert, a water type for each produced water sample in the database and training a machine-learned model with the database to predict the water type of a produced water sample given its geochemical data. The method further includes collecting a third plurality of produced water samples, performing geochemical analysis on the third plurality of produced water samples, and determining, with the trained machine-learned model, the water type for each produced water sample in the third plurality of produced water samples using the third geochemical data set.