一种基于深度学习集成模型的气象预报方法及系统
By combining CNN-LSTM, ConvLSTM and CNN3D models, the problems of high computational cost and insufficient accuracy in existing weather forecasting are solved, and efficient and accurate weather forecasting results are achieved.
CN120703867BActive Publication Date: 2026-07-17CHANGGUANG SATELLITE TECH CO LTD
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHANGGUANG SATELLITE TECH CO LTD
- Filing Date
- 2025-06-13
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing weather forecasting methods are computationally expensive when integrating multiple models, and lack accuracy, stability, and robustness with complex variables, especially in precipitation forecasting.
Method used
A weather forecasting method based on a deep learning ensemble model is adopted. By combining CNN-LSTM, ConvLSTM and CNN3D models, and using error weighting and Bayesian optimization ensemble methods, the final ensemble prediction results are generated.
Benefits of technology
It significantly improves the accuracy and stability of weather forecasts, especially in precipitation forecasts where the error is reduced by more than 60%, simplifies model design, and enhances generalization ability.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN120703867B_ABST
Abstract
本发明涉及多要素气象预报技术领域，具体涉及一种基于深度学习集成模型的气象预报方法及系统；一种基于深度学习集成模型的气象预报方法包括：获取气象数据，并进一步构建训练数据集；使用训练数据集分别训练CNN‑LSTM模型、ConvLSTM模型和CNN3D模型；各模型分别对输入各个模型的气象数据进行处理，输出3个不同的预测结果；分别采用基于误差的加权集成方法、贝叶斯优化集成方法对3个不同的预测结果进行集成预测，得到最终的集成预测结果。本发明整体方案设计简单，易于实现，且预测精度相比传统方案更高。本发明通过结合不同模型的优势，集成方法显著提升了稳定性和泛化能力，尤其在复杂变量如降水上表现更鲁棒。
Need to check novelty before this filing date? Find Prior Art