AI Emission Inversion System Using 3D CNN for Pollution Source Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies for predicting air pollution and tracing sources are limited by data dependency, computational complexity, uncertainty in chemical reactions, scale limitations, and adaptability to environmental changes, making it difficult to accurately identify pollution sources and update emission lists in a timely manner.
Innovation Solution
An inversion method using artificial intelligence and big data is developed to determine pollution source lists, which involves acquiring and preprocessing weather, emission, and concentration data, and using a 3D CNN algorithm to establish a relationship between pollutant concentration and emission, with the Integrated Gradients method to estimate the influence of emissions on concentration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If chemical transport models (CTM) are used to predict air pollution and trace sources, then detailed insight into pollutant behavior can be obtained, but computational complexity increases significantly requiring great computing resources and time
Solution Approach 1:
The patent replaces the traditional chemical transport model (CTM) with a deep learning-based neural network model. This substitution transitions from a physics-chemistry-based mechanical system to an data-driven intelligent system, maintaining prediction accuracy while significantly reducing computational complexity and resource requirements.
Solution Approach 2:
The patent changes the fundamental parameters of the modeling approach by using learned parameters from training data instead of fixed physical-chemical parameters. The deep learning model learns emission inventory parameters, meteorological parameters, and their interactions from historical data, enabling accurate predictions with lower computational overhead.
2Area of stationary object
If chemical transport models (CTM) are used for large area coverage or long-time simulation, then comprehensive pollution analysis can be achieved, but computational resources and time requirements increase greatly
Solution Approach 1:
The deep learning model replaces the computationally intensive CTM system, enabling rapid simulations across large areas and extended time periods. The trained model can process spatial and temporal variations efficiently without the exponential computational cost associated with traditional physics-based models.
Solution Approach 2:
The patent performs preliminary training of the deep learning model using historical emission inventories, meteorological data, and pollution observations. This preliminary action creates a pre-trained model that can rapidly predict pollution levels for new scenarios without requiring repeated complex computations, thus reducing computation time for large-scale and long-term simulations.
3Measurement precision
If manual monitoring methods are used to characterize air pollution, then pollution concentration can be measured, but it is difficult to trace back to sources and supervision workload becomes very heavy
Solution Approach 1:
The patent replaces manual monitoring and analysis methods with an automated deep learning-based inversion system. This intelligent system automatically traces pollution sources by analyzing the relationship between emission inventories, meteorological conditions, and observed pollution levels, eliminating the need for heavy manual supervision while improving source identification accuracy.
Solution Approach 2:
The deep learning model performs self-service by automatically learning the complex relationships between emissions, weather conditions, and pollution levels from training data. The model can independently identify pollution sources and quantify their contributions without requiring manual intervention, thus reducing supervision workload while maintaining or improving measurement precision.
Data Source
AI summary
The present disclosure provides an inversion method for determining a pollution source list based on artificial intelligence and big data, an inversion system for determining the pollution source list based on artificial intelligence and big data, and applications thereof, which provides basic data support for government sectors to formulate relevant environmental protection measures. The specific technical solution employed in the present disclosure is as follows: finding out an emission source that makes the highest contribution to the pollutant concentration of any cell with an advanced 3D CNN artificial intelligence algorithm based on artificial intelligence and big data, establishing a model of the relationship between pollutant concentration and emission, and finding out the relationship between pollutant concentration and emission with machine learning technology, i.e., estimating an emission from a given pollutant concentration, and estimating a pollutant concentration from a given emission.