基于深度学习的LAFs短临强降水预测模型研究

打开文本图片集

中图分类号：P426.6 文献标志码：A 文章编号：2095-2945(2025)14-0072-04

Abstract:Whenshort-termandimminentrainfallafectsthegenerationofrunoffandthedistributionofwaterresources, accurateforecastingcanbringhugeeconomicbenefitstorelevantdepartments.Inordertoimprovetheacuracyofheavyrainfal forecast，anLAFsshortimminentheavyrainfallforecastmethodbasedonLSTM-Atentioncombinedwiththeaumulatoris proposed.Themodelfirstusesacubicpolynomialinterpolationmethodtogridtheactualobservationelementsoftheground station;Thenthedataisextractedandfusedthroughtheaccumulator;Finaly，theobtainedfeaturefactorsareusedasinputsto themodelformodelprediction.ThreatscoreTSandmeansquareerorareselectedasindicatorstocomprehensivelyevaluatethe performanceof theproposedmodel，andcomparedwithLSTMandConvLSTM.Theresultsshowthattheperformanceof the proposed model is better than the other two models,and its TS score is 2% 业 3% higher than that of the other two models,and 5% （2 higherthantheactualoperationalforecastlevelinthesameregion,indicatingthattheproposedmodelhascertainpracticalvalue.

Keywords:longshort-termmemorynetwork (LSTM);atentionmechanism;feed-forwardnetwork；short-termimminentheavy precipitation; accumulator

短时强降水引发的山洪、城市内涝和地质灾害已屡见不鲜，精准的降水预报对于防洪、水资源管理、空中交管以及能源管理有重要的战略意义。（剩余8653字）