摘要
为科学评价云南省2006-2022年及2025年水资源空间均衡状态,建立基于社交网络搜索(SNS)算法、登山队优化(MTBO)算法优化双向门控循环单元(BIGRU)、双向长短时记忆(BILSTM)网络的水资源空间均衡评价模型。首先,从水资源支撑、水资源压力、水资源调控力3个方面遴选15个指标构建水资源空间均衡评价指标体系和等级标准,采用线性内插和随机选取的方法生成样本构建BIGRU、BILSTM适应度函数;其次,简要介绍SNS、MTBO算法原理,利用SNS、MTBO优化BiGRU、BiLSTM隐含层神经元数、学习率(超参数)构建SNS-BIGRU、MTBO-BIGRU、SNS-BILSTM、MTBOBILSTM模型,通过不同样本大小和连续10次运行的方法验证SNS-BIGRU等4种模型的稳健性;最后利用SNS-BIGRU、MTBO-BIGRU、SNS-BILSTM、MTBO-BILSTM模型对云南省2006-2022年及2025年水资源空间均衡进行评价,并与SNS-支持向量机(SVM)、MTBO-SVM和模糊综合评价法的评价结果作对比。结果表明:①所建立的SNS-BIGRU等4种模型具有较好的识别精度和稳健性能;SNS、MTBO能有效优化BIGRU、BILSTM超参数,提升BIGRU、BILSTM预测性能。②SNS-BIGRU等4种模型对云南省2006-2011年水资源空间均衡评价为“不均衡”,2012-2013年评价为“较不均衡”,2014-2018年评价为“临界均衡”,2019-2022年评价为“较均衡”,2025年基本可达到“均衡”水平;4种模型评价结果与SNSSVM、MTBO-SVM、模糊综合评价法有3年存在1个等级的差异。本文构建及提出的模型方法可为水资源空间均衡评价提供参考与借鉴。
This paper evaluates the spatial equilibrium status of water resources in Yunnan Province from 2006 to 2022 and 2025 scientifical⁃ly,and proposes a water resource spatial equilibrium evaluation model based on social network search(SNS)algorithm,mountaineering team optimization(MTBO)algorithm,bi-directional gated loop unit(BIGRU)optimization,and bi-directional long short-term memory(BILSTM)network optimization.First,15 indicators are selected from three aspects of water resources support,water resources pressure and water resources regulation to build the evaluation index system and grade standard of water resources spatial balance,and the samples are generated by linear interpolation and random selection methods to build BiGRU and BiLSTM fitness function.Secondly,the principles of SNS and MTBO algorithms are briefly introduced.SNS-BIGRU,MTBO-BIGRU,SNS-BILSTM,and MTBO-BILSTM models are construct⁃ed by using SNS and MTBO to optimize the number of neurons in the hidden layer of BiGRU and BilSTM and the learning rate(hyperparame⁃ter).The robustness of four models,including SNS-BIGRU,MTBO-BIGRU,SNS-BILSTM,and MTBO-BILSTM,is verified through dif⁃ferent sample sizes and 10 consecutive runs.Finally,SNS-BIGRU,MTBO-BIGRU,SNS-BILSTM,and MTBO-BILSTM models are used to evaluate the spatial balance of water resources in Yunnan Province from 2006 to 2022 and 2025,and the evaluation results are compared with those of SNS-Support Vector Machine(SVM),MTBO-SVM,and fuzzy comprehensive evaluation methods.The results show that:①the models such as SNS-BIGRU have good recognition accuracy and robustness.SNS and MTBO can effectively optimize the hyper-parame⁃ter of BIGRU and BILSTM,and improve the prediction performance of BIGRU and BiLSTM.②SNS-BIGRU and other four models evaluate the spatial balance of water resources in Yunnan Province from 2006 to 2011 as“unbalanced”,from 2012 to 2013 as“relatively unbal⁃anced”,from 2014 to 2018 as“critical equilibrium”,from 2019 to 2022 as“relatively balanced”,and by 2
作者
李杰
崔东文
LI Jie;CUI Dong-wen(Yunnan Institute of Water Resources and Hydropower Survey,Design and Research,Kunming 650000,Yunnan Province,China;Wenshan Water Bureau,Yunnan Province,Wenshan 663000,Yunnan Province,China)
出处
《中国农村水利水电》
北大核心
2023年第11期1-9,共9页
China Rural Water and Hydropower
基金
国家高分辨率对地观测系统重大科技专项(89-Y50-G31-9001-22/23-05)
云南省高层次人才及创新团队建设(2018HC024)。
关键词
水资源空间均衡
指标体系
双向门控循环单元
双向长短时记忆网络
社交网络搜索算法
登山队优化算法
云南省
water resources spatial equilibrium
index system
bidirectional gated recurrent unit
bidirectional long short-term memory
so⁃cial network search algorithm
mountaineering team-based optimization algorithm
Yunnan Province
