科技创新与应用2025，Vol.15Issue(14)：72-75,4.DOI:10.19981/j.CN23-1581/G3.2025.14.016
基于深度学习的LAFs短临强降水预测模型研究
摘要
Abstract
When short-term and imminent rainfall affects the generation of runoff and the distribution of water resources,accurate forecasting can bring huge economic benefits to relevant departments.In order to improve the accuracy of heavy rainfall forecast,an LAFs short-imminent heavy rainfall forecast method based on LSTM-Attention combined with the accumulator is proposed.The model first uses a cubic polynomial interpolation method to grid the actual observation elements of the ground station;Then the data is extracted and fused through the accumulator;Finally,the obtained feature factors are used as inputs to the model for model prediction.Threat score TS and mean square error are selected as indicators to comprehensively evaluate the performance of the proposed model,and compared with LSTM and ConvLSTM.The results show that the performance of the proposed model is better than the other two models,and its TS score is 2%～3%higher than that of the other two models,and 5%higher than the actual operational forecast level in the same region,indicating that the proposed model has certain practical value.关键词
长短期记忆网络/注意力机制/前馈网络/短临强降水/预提器Key words
long short-term memory network(LSTM)/attention mechanism/feed-forward network/short-term imminent heavy precipitation/accumulator分类
天文与地球科学引用本文复制引用
黄俊,白龙,卓健,李涛,谭斐..基于深度学习的LAFs短临强降水预测模型研究[J].科技创新与应用,2025,15(14):72-75,4.基金项目
广西壮族自治区气象局指令性项目(桂气科2024ZL01) （桂气科2024ZL01）
