基于物理特性自动预测雷电的集成学习方法及存储介质
By employing ensemble learning methods and combining multiple meteorological elements and physical characteristics, an ensemble learning model was constructed, which solved the problems of accuracy and reliability in short-term lightning forecasting under complex conditions, and achieved efficient lightning forecasting.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SICHUAN METEOROLOGICAL DISASTER PREVENTION TECH CENT
- Filing Date
- 2026-03-23
- Publication Date
- 2026-07-17
AI Technical Summary
Existing short-term lightning prediction methods lack generalization ability under complex meteorological conditions. Single deep learning models are unable to fully capture the complex physical processes and multi-scale characteristics of lightning occurrence, resulting in poor prediction accuracy and reliability.
An ensemble learning method for automatic lightning prediction based on physical characteristics is constructed. The ensemble learning model is trained by input data of multiple meteorological elements, including a prediction module, a base learner module, and a meta learner module. It integrates long short-term memory network and Transformer network, and combines fully convolutional neural network, U-Net network and threshold matrix of lightning-related physical features to achieve end-to-end learning for lightning prediction.
It improves the accuracy and reliability of short-term lightning prediction, enhances the model's predictive stability and physical interpretability in variable environments, and forms a hybrid prediction framework that combines the advantages of physical understanding and data mining.
Smart Images
Figure CN122047559B_ABST