ISSN 1000-0526
CN 11-2282/P
CN 11-2282/P
A Deep Learning Prediction Model for Winter Precipitation Types Based on Sequential Fusion Encoder
- Article
- Figures
- Metrics
- Preview PDF
- Reference
- Related
- Cited by
- Materials
Abstract:
This paper proposes a deep learning model based on a sequential fusion encoder, which integrates the advantages of Convolutional Neural Networks (CNN), Convolutional Gated Recurrent Units (ConvGRU), and Transformer encoders, aiming to enhance the accuracy of winter precipitation phase prediction. Utilizing hourly precipitation weather phenomenon observational data from Wuhan over a 15-year period during winters from 2010 to 2024, as well as reanalysis data from the European Centre for Medium-Range Weather Forecasts, the model effectively addresses data imbalance issues through sample processing and data augmentation. Experimental results demonstrate superior performance in predicting solid precipitation, exhibiting high reliability in forecasting sleet and snow. By validating the model against complex weather processes in February 2024, its high precision in snowfall prediction is confirmed, though limitations are revealed in rapidly changing precipitation phases. This study offers an efficient and intelligent solution for deep learning-based prediction of winter precipitation phases, laying a foundation for further model optimization and enhanced predictive capabilities.
Keywords:
Project Supported:
Wuhan 2022 Knowledge Innovation Special Project
Clc Number:
Mobile website
