【作者】 卢金锁；

【导师】 黄廷林；

【作者基本信息】 西安建筑科技大学 ， 市政工程， 2006， 博士

【摘要】 911事件后，供水系统的安全预警成为西方发达国家供水领域一个非常重要的研究方向。随着地表水厂原水水质污染恶化事件的不断发生，原水水质预警系统作为有效地保障供水安全的措施和系统，已受到国内外的普遍关注。目前虽对原水水质预警系统有一定的认识和实际应用，但仍局限于建立在水质在线监测基础上的高投入和理想化的原水水质预警系统上，同时鉴于理论的不足和与水厂实际结合不密切等缺点，在我国经济不发达的特定国情下，它在地表水厂推广应用还存在诸多障碍。本文通过分析原水污染及水质恶化事件的特点，提出了按照原水水质恶化事件类别通过不同方式进行水质预警的理论框架。针对研究案例——天津芥园水厂的原水藻类高发特点，创造性地将聚类分析与神经网络(ANN)结合，建立了预测原水中叶绿素的优化ANN模型；在建立了原水水质在线监测系统基础上，运用GPRS网络实现监测系统与水厂调度之间点对点的数据传输，开发了远程原水水质监测软件，并首次采用决策树技术建立了基于在线监测系统的水质预测模型；结合中试试验，提出了针对藻类高发的芥园水厂净水预案及预警决策框图，为芥园及相似水厂运行提供指导和建议。论文的主要成果和创新性主要体现在以下几方面：(1)完善了原水水质预警理论框架地表水源污染可分为持续性污染、周期性污染和偶发性污染三大类，分别导致持续趋势性、周期间断性、偶发瞬时性水质变化，前两类可在水质预测和水质监测基础上进行原水水质预警，而后者只能在原水水质连续在线监测的基础上，进行预警；在目前我国持续性污染和周期性污染引起的原水水质恶化没有得到很好解决之前，没有足够资金支持情况下，可充分利用多年的原水水质监测资料，针对持续趋势性水质恶化和周期间断性水质恶化，建立基于传统监测的水质预警系统；在有资金来源或偶发性污染风险较大时，可在建立原水水质在线监测系统的基础上，建立原水水质预警系统。(2)建立了基于水厂传统原水监测数据的藻类高发预测模型在收集天津芥园水厂的日常规监测数据基础上，通过对数据相关分析及箱形图分析，研究了指标之间的表针关系及各月的水质指标异常线；通过对水质监测数据的逐月频率分析，研究了藻类逐月变化规律及警戒藻类高发时间；通过对原水叶绿素变化曲线的小波分析，研究了藻类变化的年变化规律及其主要的影响因素；通过ROC曲线分析，研究了常规指标与滞后数日原水中叶绿素的关系；建立了预测第3曰原水中叶绿素神经网络模型，模型输入变量为当前水质指标：叶绿素、浊度、氨氮、水温、pH值、总碱度，通过仿真验证，得到叶绿素的预测值与实测值间的相关系数达到0.88；预测的平均准确率达到85％以上。(3)建立了基于在线监测的藻类高发预测模型在芥园水厂的西河预沉池进口处建立原水水质在线监测系统，在线监测指标有：浊度、pH、溶解氧、水温、氧化还原电位、电导率、氨氮、正磷、大地照度，通过GPRS网络，开发远程原水水质监测软件实现了监测数据的在线采集以及在线监测系统与水厂调度中心之间点对点无线数据传输，优化了系统运行，建立了基于在线监测系统的预测原水中次日叶绿素的决策树模型，经仿真验证，模型的平均预测准确率可达到80％。(4)提出了芥园水厂原水藻类高发的预案及预警决策框图在调查了芥园水厂新改造工艺和水源系统后，详细分析了水厂处理高藻水的各种可能的潜在能力；结合中试试验研究，提出了藻类高发程度不同情况下，相应的水厂及水源相关部门应对预案，结合以上成果，提出芥园水厂针对藻类高发的水质预警决策框图。更多还原

【Abstract】 Since the 911 incident, the research on security early warning of water supply system has become very important in western developed countries. The early-warning system (EWS) on raw water, being an effective safeguard technique for the water supply, has made more and more domestic and abroad scholars begin to study it as the result of series of incidents of source surface water, At present, based on the online monitoring system of raw water quality with high fund input, the research and application of EWS is still confined to idealization although there are some finite knowledge and practical applications on it. Moreover because of some obstacles such as lacking of theories and divorcing from the reality of the plant the EWS’s large scale application on source surface water of drinking water utilities is still very difficult during the special status of not developed economy in our nation.Based on the analysis of the characteristics of the source water pollution and deterioration, an innovated theory of taking different modes, according to the classification of the deteriorated water quality incident, to early warn is boldly proposed in this paper. The studied case is JIEYUAN water plants (JYWP) in Tianjin, whose key points are the algae-laden raw water and its treatment. The optimum ANN model to forecast the chlorophyll in the raw water is creatively established by the author by combining cluster analysis and artificial neural networks (ANN). The GPRS network is used to transmit point-to-point datum between the online monitoring system constructed for raw water of JYWP and the utility’s control central; And the decision tree technique is first introduced to establish a model for predicting water quality based on the on-line monitoring system; Unifying the pilot-tests on the treatment of algae-laden waters, a water purification draft scheme and a warning decision diagram were proposed aimed at the algae-laden raw waters to provide guidance and advice for JYWP and other similar water plants. The achievements and innovation of this paper are as follows: (1) The theoretical frame of the raw water quality early warning system was completed and betteredThe surface water pollution is divided into three categories: sustained, periodic and accidental pollution, which respectively lead to lasting tendency, periodic discontinuity and accidental short time water quality deterioration. The first two can be early warned on the basis of water quality forecast and monitoring, while the latter can only be early warned based on the raw water contentious online monitoring; Before the sustained pollution and the periodic pollution which cause the raw water quality deterioration has not been well resolved at present, the traditional lab-monitoring information since water works begin for many years can be taken full advantage to establish the water quality EWS in respond to the sustained and the periodic water deterioration whether there are sufficient funds budget or not. But if there are a larger source of funds or accidental pollution risks, EWSs for raw water can be built on the base of online monitoring systems for water quality.(2) A prediction model for algae of raw water was established on the base of traditional lab-monitoring datum of raw water quality.Based on the day-to-day monitoring datum from JYWP, the relationship between indexes and the unusual line on each water quality indexes every month are studied by analyzing the correlation and box-plot; the law of the algae’s changing month by month and the high level algae alert time are researched by the frequency analysis of water quality monitoring datum month by month; the law of the algae annual changes and its main influencing factors are studied through the wavelet analysis of chlorophyll changes in the source water; the relationship between the common water indexes and chlorophyll quantity in the raw water delayed of a few days is studied through the ROC curve analysis; An ANN model is created to forecast the chlorophyll in the raw water the day after tomorrow, whose input variables are the current water quality indexes: chlorophyll, turbidity, ammonia, temperature, pH, total alkalinity. Through simulation verification, the correlation coefficient of 0.86 between the predicted values and tested values of the chlorophyll is gained, and the average forecast accuracy rate is more than 85%.(3) A prediction model for algae of raw water was established on the online monitoring system for raw water qualityThe online monitoring system was build near the Xihe preliminary sedimentation inlet for raw water of JYWP, monitoring indexes as follows: turbidity, pH, dissolved oxygen, water temperature, redox potential, conductivity, ammonia nitrogen, orthophosphate, radar luminance. The point-to-point wireless data transmission between the online monitoring system and the TYWP control center is realized through the GPRS networks. After optimizing the system operation, the decision tree model for forecasting the chlorophyll of the next day in raw water is established based on the collecting datum from the online monitoring system; the average predicted accurate rate of the model can reach to 80% validating by simulation.(4) A water purification draft scheme and a warning decision diagram for the algae-laden raw water of JYWP and other similar water plants were proposed.After the investigation of the rebuilt water treatment process and the flow path of source water of JYWP, a detailed analysis about the potential of the water utility and related parts is carried out for the treatment of algae-laden water. Unifying pilot-tests, the corresponding response plans are proposed for the water plants and related sectors according to the quantity of algae in source water. Ultimately, the early-warning decision diagram against the treatment algae-laden raw water is proposed for JYWP and other similar water plants.更多还原