Autoformer-Based Model for Predicting and Assessing Wheat Quality Changes of Pesticide Residues during Storage

Area of Science:

• Agricultural Science
• Food Science
• Data Science

Background:

• Proper grain storage is crucial for maintaining food quality and safety.
• Wheat is a globally significant grain reserve due to its desirable cultivation and storage characteristics.
• Accurate assessment of stored wheat quality, particularly concerning pesticide residues, is essential.

Purpose of the Study:

• To develop an advanced model for predicting changes in wheat pesticide residue concentrations during storage.
• To establish a comprehensive wheat quality assessment index (Q) integrating predicted and actual pesticide residue data.
• To enhance the scientific foundation for improving the quality management of stored wheat.

Main Methods:

• Collected and analyzed monitoring data on storage environmental parameters and pesticide residue concentrations from over 20 regions in China.
• Developed and applied an Autoformer-based model for predicting pesticide residue concentration dynamics.
• Utilized the K-means++ algorithm in conjunction with a comprehensive wheat quality index (Q) for quality assessment.

Main Results:

• The Autoformer model achieved superior prediction performance with minimal error values.
• Key error metrics included a Mean Absolute Error (MAE) of 0.11017.
• The integrated approach effectively assessed stored wheat quality based on pesticide residue levels.

Conclusions:

• The Autoformer model provides a robust and accurate method for predicting pesticide residue changes in stored wheat.
• The developed quality assessment framework offers practical technical support for enhancing stored wheat quality.
• This research contributes to ensuring the safety and quality of global wheat reserves through improved storage management.