Machine Learning Methane Emission Monitoring Under Turbulent Dispersion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methane emission monitoring technologies face challenges in accurately linking sensor output data to emission rates due to the complex turbulent flow of air in the atmospheric boundary layer, leading to inaccuracies in simplified models and the intractability of full numerical simulations.
Innovation Solution
A machine learning-based approach that combines a regression model and a classification model, utilizing physics-based models to augment sensor data and generate predictions of emission rates and presence, leveraging artificial neural networks for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If simplified statistical models such as the Gaussian plume model are used to capture average behavior in steady wind, then the model complexity is reduced and ease of operation is improved, but measurement precision and reliability of emission quantification deteriorate because the real atmosphere does not satisfy the model assumptions
Solution Approach 1:
The patent introduces machine learning models as an intermediary between sensor measurements and emission rate predictions. Instead of directly applying simplified physics models that assume steady wind and Gaussian dispersion, the ML models learn the complex nonlinear relationships from training data, acting as a mediator that translates sensor readings into accurate emission estimates without requiring the atmosphere to satisfy strict model assumptions
Solution Approach 2:
The patent transforms the approach by changing from fixed physics-based parameters to adaptive data-driven parameters. The system uses machine learning models that can dynamically adjust to varying atmospheric conditions, replacing the static parameters of simplified models with flexible, condition-dependent predictions that maintain precision across diverse real-world scenarios
2Measurement precision
If full numerical simulation of the complete Navier-Stokes equations is performed, then measurement precision and reliability of emission quantification are improved, but device complexity and computational requirements become intractable for real full-size open-air facilities
Solution Approach 1:
The patent segments the complex fluid dynamics problem into manageable components by using machine learning models that capture essential dispersion patterns without simulating every turbulent eddy. The ML approach divides the computational task into training phase (where complex patterns are learned from data) and prediction phase (where simple inference is performed), making the system tractable for real facilities
Solution Approach 2:
The patent creates simplified representations of complex atmospheric processes through machine learning models trained on comprehensive data. Instead of directly simulating full Navier-Stokes equations in real-time, the system uses trained ML models that copy the essential behavior of complex flows in a computationally efficient manner, providing accurate predictions without the intractable complexity of full numerical simulation
3Quantity of substance
If low-cost methane sensors are deployed to detect emissions, then cost is reduced and accessibility is improved, but the ability to accurately link sensor output to emission rates deteriorates due to complex turbulent flow and lack of appropriate data for modeling
Solution Approach 1:
The patent performs preliminary action by training machine learning models on comprehensive datasets that capture complex turbulent flow patterns and atmospheric conditions before actual emission monitoring begins. This pre-training phase creates a knowledge base that enables the low-cost sensor system to accurately predict emission rates without requiring complex hardware or real-time complex simulations
Solution Approach 2:
The patent substitutes the mechanical/physical complexity of precise emission measurement systems with an information-processing approach. Instead of using expensive, complex sensor systems or full numerical simulations, the invention replaces the physical complexity with machine learning algorithms that process sensor data to infer emission rates, maintaining precision while using affordable sensors
Data Source
AI summary
A method implements machine learning based methane emissions monitoring. The method includes collecting sensor data from a plurality of sensors. The method further includes applying an augmentation model to the sensor data to form a regression training set. The method further includes creating a classification training set for a classification model by replacing regression output values from the regression training set with classification output values. The classification output values include binary values. The method further includes training the regression model with the regression training set to generate a regression prediction. The method further includes training the classification model with the classification training set to generate a classification prediction.