北方农牧交错带干旱高精度估算模型构建  

Construction of High-precision Drought Estimation Model in the Northern Agro-pastoral Ecotone

作　　者：刘洪伟 李鹏程[2,3] 张敏 孙燕飞 LIU Hong-wei;LI Peng-cheng;ZHANG Min;SUN Yan-fei(Shandong Provincial Water Resources Survey and Design Institute Co.,Ltd,Jinan 250013,Shandong Province,China;School of Geographic Sciences,Hebei Normal University,Shijiazhuang 050010,Hebei Province,China;Hebei University of Water Resources and Electric Engineering,Cangzhou 061001,Hebei Province,China;Zhaoyuan Jindu Water Conservancy Survey and Design Co.,Ltd,Yantai 265400,Shandong Province,China)

机构地区：[1]山东省水利勘测设计院有限公司,山东济南250013 [2]河北师范大学地理科学学院,河北石家庄050010 [3]河北水利电力学院,河北沧州061001 [4]招远市金都水利勘测设计有限公司,山东烟台265400

出　　处：《节水灌溉》2025年第2期53-61,共9页Water Saving Irrigation

基　　金：河北省教育厅科学研究项目资助(BJ2021056);河北水利电力学院河北省高等学校基本科研业务费研究项目(202004)。

摘　　要：为构建区域干旱的高精度简化估算模型,以中国北方农牧交错带为研究区域,选择该区域12个气象站点,计算不同站点3个月、6个月、12个月的标准化降雨蒸散指数(SPEI-3、SPEI-6、SPEI-12),以表征区域干旱,使用时间卷积神经网络模型(TCN)来提取序列数据的特征,同时输入到双向长短期记忆神经网络模型(BiLSTM)中进行进一步的处理,构建组合模型(BT),采用麻雀搜索算法(SSA)和Attention机制对组合模型进行优化,构建SSA-BiLSTM-TCN-Attention模型(SBTA),同时计算了SBTA模型精度,基于均方误差(MSE)、决定系数(R^(2))和效率系数(E_(NS))以及GPI指数的模型精度评价体系进行精度验证,结果表明:SBTA模型MSE值仅为0.041~0.200,R^(2)和E_(NS)在全区取值均在0.9以上,在全区的误差最低、一致性最高,在所有模型中精度排名第1,可推荐SBTA模型用于北方农牧交错带干旱估算当中。In order to construct a high-precision and simplified prediction model for regional drought,in this paper we selected 12 meteorological stations in the northern agro-pastoral ecotone as the research area,and calculated the standardized rainfall evapotranspiration index(SPEI-3,SPEI-6,SPEI-12)of different stations for 3 months,6 months,12 months to characterize regional drought.The combined model was constructed based on BiLSTM and TCN,and the Sparrow search algorithm(SSA)and the Attention mechanism were used to optimize the combined model.The SSA-BILSTM-TCN-Attention model(SBTA)was constructed.At the same time,the accuracy of SBTA model was calculated,constructing a model accuracy evaluation system based on the mean square error(MSE),the coefficient of determination(R^(2))and the efficiency coefficient(E_(NS)),and the GPI index.The results showed that the MSE value of SBTA model was only 0.041~0.200,both R^(2) and E_(NS) took values above 0.9 in the whole region,which exhibited the lowest error and the highest consistency in the whole region,and the accuracy ranked first among all models.The SBTA model can be recommended for regional drought prediction.

关 键 词：干旱估算模型 北方农牧交错带 标准化降雨蒸散指数 组合模型 麻雀搜索算法 Attention机制 

分 类 号：P426[天文地球—大气科学及气象学]