基于人工神经网络的流域多主体水生态补偿核算模型  

Multi-agent Water Ecological Compensation Accounting Model Based on Artificial Neural Network

作　　者：刘家宏[1,2] 鲁佳慧 燕文昌 于赢东[1] 张蕊 LIU Jiahong;LU Jiahui;YAN Wenchang;YU Yingdong;ZHANG Rui(China Institute of Water Resources and Hydropower Research,State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin,China,Beijing 100038,China;Engineering and Technology Research Center for Water resources and Hydroecology of the Ministry of Water Resources,Beijing 100044,China;Shanxi Water Resources Research Institute,Taiyuan 030000,China)

机构地区：[1]中国水利水电科学研究院,流域水循环模拟与调控国家重点实验室,北京100038 [2]水利部水资源与水生态工程技术研究中心,北京100044 [3]山西水资源研究所有限公司,山西太原030000

出　　处：《应用基础与工程科学学报》2022年第2期257-267,共11页Journal of Basic Science and Engineering

基　　金：国家自然科学基金项目(51739011,51879274);国家重点研发计划项目(2016YFC0401401,2018YFC1508203)。

摘　　要：水生态补偿是协调流域内生态保护主体与经济发展主体之间竞争用水关系的重要手段.水生态补偿机制实施过程中存在补偿金量化、多主体之间额度分配比例确定等一系列难题,目前国内3个主体以上的流域水生态补偿案例较少.本文基于人工神经网络多元非线性回归方法,构建了流域多主体水生态系统服务价值的核算模型;考虑流域内各主体的人均生态服务价值的水平及其横向对比,利用S型Pearl生长曲线函数确定各主体的人均水生态补偿系数,进而计算出各主体的水生态补偿总额.模型在汾河流域进行了实证应用,基于汾河流域6个地级行政分区2013~2017年的水资源、水生态和社会经济等模拟计算了各地市逐年的生态补偿金额.结果表明,2013年忻州市和晋中市受偿金额分别为1185万元和45万元,太原市、临汾市、吕梁市和运城市的支付金额分别为423万元、393万元、291万元和123万元.2013~2017年,太原市水生态补偿一直处于支付状态,其中2015年支付所占比例最高,为59%,2016~2017年支付金额减少,2017年支付比例降低为6%.基于人工神经网络流域多主体水生态补偿核算模型可实现补偿金的动态核算,为流域补偿提供借鉴和参考.Water ecological compensation is one of the effective countermeasures to coordinate the relationship between competitive water users,ecological protection and economic development.During the implementation of the water ecological compensation,there are a series of difficulties such as quantifying compensation funds and determining the quota allocation ratio among multiple agents,which leads to few successful cases of water ecology compensation in river basins with more than three agents in China.Based on the multivariate nonlinear regression method of artificial neural network,an accounting model of water ecosystem service value was constructed for multi-agent.Considering the level of per capita ecological service value of each agent and horizontal comparison,the S-shaped Pearl growth curve function was used to calculate the coefficient and the total compensation amount of each subject.The model was applied in the Fenhe River basin.Based on the water resources,ecology,and economy of the six administrative subdistricts in basin from 2013 to 2017,the annual ecological compensation amount of each agent was calculated.The results showed that in 2013,Xinzhou City and Jinzhong City received 11.85 million yuan and 0.45 million yuan in compensation,and Taiyuan,Linfen,Lüliang,and Yuncheng paid 4.23 million,3.39 million,2.91 million and 1.23 million,respectively.From 2013 to 2017,the highest proportion of payments in Taiyuan was 59%,and the proportion of payments in 2017 decreased to 6%.The artificial neural network-based multi-agent ecological compensation accounting model can quantify the amount of compensation yearly,which provides the key reference for river basin compensation.

关 键 词：水生态补偿 核算模型 人工神经网络 水生态系统服务价值 汾河流域 

分 类 号：X-1[环境科学与工程]