Carbon Dioxide Removal
AI/ML Approaches for Research and Development Application
Approaches
Use machine learning approaches based on known labels/expectations, usually for
- Classification (e.g., Logistic, Support Vector Machine)
- Regression (e.g., Linear, Kernel Ridge)
Potential Problems
- Leverage data to better understand and predict conditions, limitation and process requirements of materials for lower-partial-pressure DAC systems.
- Predict the optimum sorbent, solvent or other material and air contractor design.
- Predict potential impacts of precipitation, air pollutants, etc.
- Forecast material life and wear
- Setting acceptable operating ranges for each system and under which conditions.
- Predict system needs for sites and overall best sites.
- Digitize and characterize legacy data for NLP models using OCR.
- Predict the impacts of various feedstock mixes/biomass types on pollution control equipment.
- Predict chemical, biological and geological system impacts over time.
Data/Input Formats
- Structured numeric data, text, image, maps, sequences, time-series data
Applications
- Predictive and prescriptive DAC methods for operation and management
- Potential impacts of weather.
- Best site locations
- Predict geological, chemical and biological systems over time.
- Real-time prediction and monitoring
- Site management and planning
Approaches
Use machine learning approaches without known labels/expectations, usually for
- Clustering
- Dimension reduction
Potential Problems
- Select the most promising materials for testing.
- Select key factors in minerals.
- Identify the most valuable data
- Determine site-specific co-benefits at potential deployment locations.
- Help locate minerals of interest.
Data/Input Formats
- Structured or unstructured data
Applications
- Identify key factors to optimize critical properties at scale for local conditions.
- Identify the most valuable new types of data that should be tracked.
- Locate minerals of interest.
Approaches
Use complex neural network architecture for
- Image and spatial classification
- Time series regression
- Natural language processing
Potential Problems
- Adjust the technology to variable boundary conditions.
- Accelerate wholistic simulations for different priority rankings, summarize new outcomes and assist in weighing the impacts.
- Detailed model to generate rapid edge processing to predict DAC load following at the local level.
- Predict impacts of weather variations on DAC systems based on existing trends.
- Suggest strategies to increase the rate/capacity of carbon uptake by the mineral types present.
- Scan data sources to identify patterns to explore the complex relationships for Ocean CDR
- Image based advanced quantification and detection.
Data/Input Formats
- Structured numeric data, text, image, maps, sequences, time-series data
Applications
- Predict impacts of weather variations
- Predictive and prescriptive operation optimization
- Develop strategies to increase rate/capacity of carbon uptake
- Image based quantification and detection
- Multi-scale and resolution prediction
Approaches
Combine more than one machine learning models to handle the tasks
- Random Forest
- XGBoost
- Committee Machine
Potential Problems
- Identify the predictable effects and setting acceptable range of operating parameters to optimize overall operations.
- Determine most cost effective technology for a specific site to include multiple considerations (geological, capital, prediction, land, water, social)
- Prioritize resources and methods to clarify the impacts of temperature, rainfall, humidity, plant-mineral interactions and similar.
- Predict and quantify social and economic impacts & spatio-temporal questions.
- Multi-dimension scale predictions.
Data/Input Formats
- Structured numeric data, text, image, maps, sequences, time-series data
Applications
- Predictive and prescriptive operation optimization
- Determine cost effective technology
- Predictive decision support
Approaches
Integrate scientific equations and knowledge into machine learning algorithms (usually using Deep Learning), such as Physics-Informed Machine Learning
Potential Problems
- Aid in scaling DAC material technology that performs well in the lab to variable conditions in the real world.
- Create digital twin to mimic testing facilities’ environment to enable virtual/hybrid experiments.
- Improve designs for gas-liquid contactors, electrode materials and membrane contactors for large-scale applications.
- Evaluate and predict complex in situ mineral stability in the presence of water & natural geochemical processes.
Data/Input Formats
- Structured numeric data, text, image, maps, sequences, time-series data
- Scientific equations, field equations, partial differential equations
Applications
- Predictive and prescriptive operation optimization
- Multi-scale and resolution prediction
- Predictive decision support