基于随机森林模型的需水预测模型及其应用

为解决需水预测模型精度问题,尝试基于随机森林模型的分类和回归功能构建需水预测模型。以苏州市需水量预测为研究实例,首先应用随机森林模型的分类功能将需水预测因子分类,经计算发现第一产业比例、人口、灌溉面积、万元产值用水量和国民经济生产总值为最重要的解释变量。在此基础上,用随机森林模型的回归功能对需水进行预测,同时采用相同的训练数据建立基于BP神经网络和RBF神经网络的需水预测模型,通过对比3个模型的预测结果,发现随机森林模型能有效预测需水量,且精度较高。     

城市水资源需求管理信息系统初探——以北京市公共生活用水为例

论述了城市水资源需求管理的发展历程和城市水资源需求管理的内涵,提出了建立城市水资源需求管理信息系统的构架体系。以北京市公共生活用水的定额管理实践为例,初步建立了以MIS/GIS为基础的北京市公共生活用水需求管理信息系统平台。系统由数据编辑、用水数据与空间数据的查询、多种统计分析、用水定额模型、需水预测模型、节水评价模型、水价敏感度模型、报表输出等主要功能组成。系统界面友好,多种报表输出,预测数据易于输出到商业化软件中,数据统计方便,系统扩展性好,可以为北京市水资源规划、管理和配置乃至社会经济发展提供科学的决策依据。     

宁夏社会经济及生态需水量预测

利用定额法、趋势预测法等不同的研究方法对宁夏的生活、生产及生态需水量进行了预测,确定了宁夏地区未来不同水平年、不同降水频率下的需水总量。结果表明:宁夏生活需水量与第二、三产业需水量呈增长趋势,但是由于农业需水量下降较大,使宁夏总需水量呈下降趋势。需水量预测结果为宁夏的水资源配置和产业结构调整提供了基础依据。     

作物腾发量实时预报与田间试验验证

获得实时、有效的作物腾发量数据是进行实时灌溉预报的基础和难点.本文根据公共的天气预报信息,提出了实时的估算作物腾发量的方法流程,并利用作物根区土壤水量平衡方程,用冬小麦田间灌溉试验的土壤水分数据对该模型和方法进行了检验;同时为检验实时参照腾发量估算方法的可靠性和实用性,对当地媒体发布的实际天气预报信息与实际气象观测值进行了对比.结果表明,在现有天气预报准确度条件下,该方法可以有效的实时估算作物腾发量,为田间实时灌溉管理与决策提供较为可靠的参数.     

Incorporating Future Climatic and Socioeconomic Variables in Water Demand Forecasting: A Case Study in Bangkok

With concerns relating to climate change, and its impacts on water supply, there is an increasing emphasis on water utilities to prepare for the anticipated changes so as to ensure sustainability in supply. Forecasting the water demand, which is done through a variety of techniques using diverse explanatory variables, is the primary requirement for any planning and management measure. However, hitherto, the use of future climatic variables in forecasting the water demand has largely been unexplored. To plug this knowledge gap, this study endeavored to forecast the water demand for the Metropolitan Waterworks Authority (MWA) in Thailand using future climatic and socioeconomic data. Accordingly, downscaled climate data from HadCM3 and extrapolated data of socioeconomic variables was used in the model development, using Artificial Neural Networks (ANN). The water demand was forecasted at two scales: annual and monthly, up to the year 2030, with good prediction accuracy (AAREs: 4.76 and 4.82 % respectively). Sensitivity analysis of the explanatory variables revealed that climatic variables have very little effect on the annual water demand. However, the monthly demand is significantly affected by climatic variables, and subsequently climate change, confirming the notion that climate change is a major constraint in ensuring water security for the future. Because the monthly water demand is used in designing storage components of the supply system, and planning inter-basin transfers if required, the results of this study provide the MWA with a useful reference for designing the water supply plan for the years ahead.     

基于Shapley值的西安市生活需水量组合预测研究

利用等维灰数递补法、非线性回归、广义回归神经网络等3种方法进行了西安市生活需水量预测,比较了不同方法的预测误差,然后根据各种方法对总预测值的信息贡献能力形成Shapley值组合需水量预测方法,计算了不同预测方法的Shapley值及其组合权重,形成组合预测需水量模型。预测结果表明:组合方法误差曲线平缓、平均误差值较小,具有一定的预测精度,适用于需水量的中短期预测。     

包头市市区居民生活用水量预测分析

用水量预测对区域水资源规划、利用和管理提供重要依据.运用灰色关联度分析法分析包头市市区居民生活用水量影响因素的基础上,分别建立多元线性回归模型、灰色GM(1,1)模型及灰色线性组合模型对该地区2009年和2010年的生活用水量进行预测分析,同时比较了三个模型的预测精度.结果表明:城市居民生活用水量与城市用水人口、人均居住面积和水价的关联度较高;2009年和2010年用水量的预测采用组合灰色模型精度最高,相对误差分别为13.6％％和6.5％,均方根相对误差为10.7％.组合预测模型的预测精度明显优于单一模型,使结果更加准确、合理,符合实际情况.     

集对分析聚类预测法在城市生活需水量预测中的应用

集对分析聚类预测法结合了集对分析中的同异反模式识别的"择近原则"和聚类分析的基本思想进行分类预测。基于该方法给出了城市生活需水量的详细方法预测,在分析过程中考虑了城市生活需水量与其影响因素之间的关系,利用城市生活用水总量和影响因素的历史数据,建立了城市生活需水量聚类预测的模型,并利用我国北京市的实际数据进行了分析计算。结果表明,该预测方法应用于城市生活需水量预测思路清晰,计算简单且精度较高,该预测方法具有较高的实用价值。     

基于二级修正的区域水资源需求量预测模型

根据区域社会经济发展的用水现状以及水资源需求的总体变化趋势,将Logistic生长曲线模型和GM(1,1)等维新息模型相结合,建立组合预测模型,组合预测区域水资源需求量,并利用ARIMA模型对区域水资源需求量的组合预测结果进行一级修正;在此基础上,根据区域节水规划的总体目标,剖析区域社会经济发展综合节水潜力,对区域水资源需求量的一级修正结果进行二级修正,最终得到区域水资源需求量的综合预测结果。算例分析结果表明,模型具有较高的预测精度。     

等维新息径向基神经网络模型在需水预测中的应用

针对现有需水预测模型进行多周期预测时存在误差随预测周期延长而累积、抗随机因素干扰能力不足等问题,提出动态等维新息径向基神经网络模型,采用聚类方法进行网络学习,并将其应用于东莞市年需水量预测中。结果表明:动态等维新息径向基神经网络模型相对于基本径向基神经网络模型具有更高的预测精度,并且预测误差不会随着预测周期的延长而累积。     

济南市水资源环境与经济发展耦合关联性测度及前景预测

结合环境库兹涅茨曲线(EKC)与耦合关联度模型对济南市水资源环境与经济发展间的耦合作用机理进行定性与定量研究,并利用弹性预测法对二者间的发展前景作出预测。结果表明:2005—2017年间水资源环境与经济发展存在耦合关系,且现状条件下生活用水、排污量与经济发展的耦合作用要强于工业用水。此外,结合EKC走势和预测结果,引入潜在风险控制指标和重点监控指标的概念,其中生活COD、工业NH₃-N为潜在风险控制指标;重点监控指标分别为生活NH₃-N、生活废水排放量及工业、生活用水量。研究成果可为促进济南市经济增长与水资源环境协调发展提供依据。     

RBF与GRNN神经网络模型在城市需水预测中的应用

针对需水量预测具有受诸多因素影响的复杂、高维和非线性等特性,本文基于RBF与GRNN神经网络算法原理,构建RBF与GRNN神经网络需水预测模型,将模型应用于城市需水预测中,并与基本BP神经网络模型以及灰色GM(1,1)需水预测模型的拟合、预测结果进行了对比分析.结果表明:①RBF与GRNN神经网络模型有着较高的拟合、预测精度,平均相对误差均在5％以内,表明研究建立的RBF与GRNN神经网络模型应用于需水预测是合理可行的,模型泛化能力强,预测精度高,算法稳定,与基本BP网络算法相比,RBF与GRNN网络模型还具有收敛速度快、调整参数少和不易陷入局部极小值等优点,可以更快地预测网络,有着良好的应用前景.②相对而言,RBF与GRNN神经网络模型预测精度要优于基本BP网络和灰色GM(1,1)模型.     

气候变化背景下流域生态需水预估−以好溪流域为例

为探究未来气候变化对流域生态需水量的影响,保障河流生态需水量,针对好溪流域进行生态需水量计算 及预测。基于好溪流域气象数据及下垫面条件建立流域生态需水模型,并根据 GF1-WFV 遥感影像数据订正后的 地表反射率和作物种植结构提升模型模拟精度。选择 CanESM2气候模式下的 RCP2.6、RCP4.5 和 RCP8.5 这 3 种 排放情景,建立气候变化背景下流域生态需水预测方法,计算现状年并预测未来年份的生态需水量及生态需水保 障程度。结果表明,基于光学遥感影像进行数据订正后,模型模拟精度有所提升,率定期的模型精度 R2从 0.80 提 升为 0.85,验证期的 R2从 0.75 提升至 0.78。应用提升精度后的模型进行生态需水预测,在 RCP2.6、RCP4.5 和 RCP8.5 情景下,2025—2100 年的年均生态需水分别增加了 0.27 亿、0.21 亿和 0.29 亿 m3,其中 RCP8.5 情景下生态 需水保障程度最高,RCP4.5 情景下生态需水保障程度最低。     

长江流域水资源预测技术

水资源预测作为现有水文、气象预报的一种拓展,必须以现有水文、气象预报手段为基础,充分利用现有的信息化技术,从气候、地理、水文水资源、水环境保护等不同领域对影响水资源变化的诸多因素进行综合分析,构建水资源预测信息平台,加强对气候、水资源等方面的基础性分析研究,开展短期气候预测技术应用研究.开展长江流域水资源预测还需要结合流域内经济发展指标,对长江流域可供水资源、工农业用水、居民生活用水、生态环境用水需求量、水质监测等信息进行收集整理和分析,通过对长江流域或流域内某一地区未来降雨量的预测,对该地区可能形成的地表径流量、水质等作进一步预测,并分析其与用水需求量之间的基本规律和经验关系,建立适合长江流域水资源预测模型,并对流域内重点区域水资源质、量在未来一定时间尺度内的变化趋势进行预测,建立相应的水资源供需关系的预警机制.目的是更好地合理利用水资源、优化水资源配置,为长江流域经济和社会的可持续发展提供强有力的技术支撑.     

基于ELES模型的城镇居民生活用水水价支付能力研究——以北京市为例

本文从支付能力的角度,根据计量经济学的需求函数理论,以我国城镇居民消费支出结构数据为基础,提出了应用扩展线性支出系统(ELES)模型,分析城镇居民生活用水水价支付能力。以北京市为例,采用1998～2004年截面数据资料,计算分析了不同时期居民用水需求与水费支出水平、居民用水的边际消费倾向、需求收入弹性以及不同收入户的用水基本需求等,以此定量研究了北京市城镇居民生活用水水价的支付能力。研究结果表明:平均收入水平下,居民对水价具有一定的支付能力,其基本用水需求均能得到满足;可支配收入增长对居民用水需求的影响不大;北京市生活用水水价还有上调的空间,但低收入户的水价支付能力有限,是重点考虑的对象。     

Mapping Urban Water Demands Using Multi-Criteria Analysis and GIS

A multi-criteria spatial data analysis method is presented to seek and model major determinants of future growth of urban water demands based on the urban growth trend by a case study performed in the city of Mytilene (Lesvos island) in north-eastern Aegean archipelagos, Greece. Domestic water use/demand is a complex element of socio-economic and physical characteristics, urban planning strategies, infrastructures and public water policies. Under a GIS environment, these factors correspond to different thematic layers, such as road network distance, distance from the city centre, distance from the coastline, topographic slope, land use/land cover, General Urban Plan of Mytilene, areas in which buildings with less than two storeys are allowed, the respective areas of buildings with more than two storeys, present population density and existing water supply and sewerage system. The Analytical Hierarchy Process applied in the evaluation of these factors led to spatial potential water demand map, which was classified into five priority zones. Results showed that a large percentage of the study area was identified as very high and high priority areas for future domestic water demand. The applied method was verified against the prepared maps showing the varying rates of water consumption and active water connections. The comparative evaluation of these maps validated the accuracy of the proposed method which can be used by the local officials of Mytilene in adopting policies and strategies aiming towards sustainable water resource management and development.     

Statistical Domestic Water Demand Model for the West Bank

Abstract

The water balance in the West Bank shows a severe deficit. Scenarios and strategies are formulated in order to overcome the deficit problem. These include options for better management of the existing water resources and the enhancement of new resources. This paper focuses on demand modeling as one of the key issues for effective water management. Although past literature about demand modeling is comparatively rich for different regions in the world, this research provides a unique study, due to the past political situation in the Middle East, for water demand modeling in the West Bank. The developed statistical domestic water demand model will assess the factors which influence domestic water use, and determine the parameters that may help in demand management. Rammallah City is used as a case study to illustrate the proposed framework of the analysis. The developed model indicates that water utility authorities can use price as a tool to ration water or encourage reduced water consumption in households.     

Determinants of Small-Area Water Consumption for the City of Phoenix,Arizona

Rapid population growth in the face of an uncertain climate future challenges the desert city of Phoenix, Arizona to consume water more prudently. To better understand the demand side of this important issue, we identified the determinants of water consumption for detached single-family residential units using ordinary least squares regression (OLS). We compared the results from the OLS model to those of a geographically weighted regression (GWR) model to determine whether there are spatial effects above and beyond the effects of the OLS variables. Determinants of residential water demand reflect both indoor and outdoor use and include household size, the presence of swimming pools, lot size, and the prevalence of landscaping that requires a moist environment. Results confirm the statistical significance of household size, the presence of a pool, landscaping practices, and lot size. Improvement of the GWR over the OLS model suggests that there are spatial effects above and beyond the effects for household size and pools – two of the four determinants of water demand. This means that census tracts exhibit water consumption behavior similar to neighboring tracts for these two variables. Model parameters can be used to investigate the effects of policies designed to regulate lot size, pool construction, and landscaping practices on water consumption and to forecast water demand in areas of new construction.     

Urban Water Demand for Service and Industrial Use: The Case of Zaragoza

Increased efforts to improve urban water management are focused on demand side policies, seeking to affect the behavior of users so that a “reasonable” use of water resources is reached. In this framework the accurate characterization of water demand play a major role in obtaining sufficient knowledge about this behavioral response to changes in price. In this paper we focus on the water demand of the services and industries connected to the public water network. To this end, we carry out an empirical estimation of urban water demand for service and industrial use in Zaragoza (Spain). The proposed model is a Koyck flow adjustment demand model, and a price specification, which is constructed as a function of the lagged average price, current marginal price and a price perception parameter. We use a dynamic panel data methodology to estimate the water demand function. As far as we are aware, this approach to service and industrial urban water demand is new in the literature. The analysis suggests that although price has a negative relationship with consumption, such an effect is reduced given that the price elasticity is lower than one in absolute value. Another relevant finding is that service and industrial urban users think that they pay a lower price than the actual price they pay.     

福建省社会经济转型期水资源供需平衡趋势分析

为了预测处于社会经济转型期的福建省各地市水资源供需平衡能力发展趋势,构建了水资源供需平衡预测指标体系。采用情景分析法、非线性回归法、自动回归法对各指标变量进行预测,采用AHP赋权、加权求和法分析福建省各地市水资源供需平衡态势。结果表明,除漳州外,各地市水资源供需平衡水平趋强,但增速减缓。由于水资源丰度、供水能力、用水效率、需水趋势各领域指数对水资源供需平衡水平贡献的变化,应注意因地因时调整水资源供需对策。