Deep Learning Fusion Model for Meteorological Data Resolution

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The inconsistency in spatial resolution and systematic errors among multi-source meteorological data hinders effective fusion and analysis, leading to information loss and poor results in local areas.

Innovation Solution

A deep learning-based meteorological big data fusion method utilizing a residual dense network (RDN) with deformable convolution and a spatial-temporal attention module to optimize super-resolution and feature extraction, ensuring high-resolution data fusion while retaining information and highlighting local and temporal features.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional average method is used to fuse multi-source meteorological data, then overall results are superior to single model, but local area effects are poor and information loss occurs due to resolution differences

Engineering Contradiction:

Improveoverall fusion accuracyVSAvoidlocal area information loss

Core Design Contradiction:

Measurement precisionVSLoss of information

Solution Approach 1:

The patent applies local quality by using spatial attention mechanisms that assign different weights to different spatial locations in the data fusion process. This allows the model to focus on locally important features while maintaining overall fusion accuracy, thereby resolving the contradiction between overall accuracy and local information preservation.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent introduces temporal dimensions and attention mechanisms that operate in additional dimensional spaces. By transforming the fusion process from simple spatial averaging to multi-dimensional attention-based fusion, the model can preserve local information while maintaining overall accuracy.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

2Area of stationary object

If multi-source meteorological data with different resolutions are used, then data coverage is improved, but resolution inconsistency and systematic errors increase

Engineering Contradiction:

Improvedata coverage areaVSAvoiddata resolution consistency

Core Design Contradiction:

Area of stationary objectVSManufacturing precision

Solution Approach 1:

The patent applies parameter changes by dynamically adjusting the resolution and weighting parameters of different data sources based on their quality and relevance. The attention mechanism automatically modulates the contribution of each data source, transforming fixed-resolution inputs into adaptively scaled representations that maintain consistency while preserving coverage.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces dynamic resolution adjustment through the attention mechanism, which adaptively determines the effective resolution contribution of each data source based on local conditions. This dynamic approach allows the system to maintain resolution consistency while utilizing multi-source data with originally different resolutions.

Inventive Principle:

Principle #15Dynamics

3Manufacturing precision

If deep learning super-resolution model is used to process multi-source meteorological data, then high-resolution fusion is achieved, but computational complexity increases

Engineering Contradiction:

Improvefusion data resolutionVSAvoidmodel computational complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent applies segmentation by dividing the complex deep learning model into modular components: feature extraction modules, attention mechanisms, and fusion modules. This segmentation allows for more efficient computation and training while achieving high-resolution fusion, reducing the overall computational complexity through structured design.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies partial action by implementing selective feature processing through attention mechanisms. Instead of processing all features at full resolution throughout the network, the model applies computational resources selectively to the most important features and regions, achieving high-resolution output with reduced overall computational complexity.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS11836605B2Meteorological big data fusion method based on deep learning

Publication Date: 2023.12.05 NANJING UNIV OF INFORMATION SCI & TECH

AI summary

The present disclosure provides a meteorological big data fusion method based on deep learning, including the following steps: constructing multi-source meteorological data samples; according to an original resolution of different climate variables, selecting a corresponding super-resolution multiple to obtain an optimized super-resolution module under the constraint of maximizing information retention efficiency; constructing a spatial-temporal attention module using a focused attention mechanism, and selecting a corresponding time stride according to periodic characteristics of different climate variables; constructing a meteorological data fusion model in combination with the optimized super-resolution model and the spatial-temporal attention module; taking a minimum resolution of climate variables as a loss function, and training the meteorological data fusion model with the multi-source meteorological data samples; and importing the acquired real-time meteorological data from multiple data sources into the trained meteorological data fusion model to obtain high-resolution fused meteorological data.