【作者】 宋筱轩；

【导师】 周泽魁； 张光新；

【作者基本信息】 浙江大学 ， 控制科学与工程， 2014， 博士

【摘要】 城市供水水源污染事件时有发生,严重时会威胁到群众的饮水安全和国家的经济发展。面对这一问题,城市供水主管部门迫切需要多环节水质预警信息支持。尤其是突发性水污染事故具有事发突然、不确定因素多等特点,快速准确地预测水质演化趋势一直是研究的难点和热点。基于此背景,本文结合国家科技重大专项课题和国家自然科学基金项目,围绕河流突发性水污染事故预警方法和系统实现技术开展理论与应用研究,主要工作和创新点如下：(1)研究提出了动态数据驱动的河流突发性水污染事故预警系统架构,为河流突发性水污染事故预警研究提供了新的研究方法和思路。完成了该架构的总体设计,分析了基于该架构下水质预警系统的工作原理,与基于通常架构下的水质预警系统工作流程进行了比较,剖析了动态数据驱动范式下的系统架构具备多模型集成、共生反馈、动态性和自学习性等特点和优势。(2)针对目前水质预警模型在动态适应性和应用迅捷性等方面急需提升等问题,研究了基于动态数据驱动的河流突发性水污染事故预警系统架构下的水质动态预警方法。在初始预警模型基础上,结合历史经验数据和动态监测数据,首先采用基于案例的推理方法进行初始校正,实现初始预警；然后,针对水文数据和边界条件数据较缺乏等情况下的精确化预测问题,研究了应用后续实测数据和改进的网格寻优算法对水质预测预警模型参数进行自适应校正的技术,针对预警模型往往因模型概化而导致结构性误差等问题,进一步研究提出了基于时间序列算法的预测预警结果动态校正方法。以实际发生的河流突发性水污染事故为应用案例,对所研究动态预警方法进行了应用研究与分析。应用结果表明该方法可有效地对河流突发性水污染事故进行预警,预测结果与实测结果之间经动态校正而减小偏差。(3)针对应急监测点布局缺乏理论指导、应急部门难以快速合理选点等问题,研究了动态数据驱动的河流突发性水污染事故预警系统架构下的应急监测点选取、评价和布局优化方法,建立了基于模糊层次分析法的应急监测点布局技术方案,为水质事件不同阶段应急监测点敏感性和重要性的量化评估提供了技术支撑,实现了有限应急监测力量的合理、及时和动态调配。该技术的研究国内外尚鲜有报道。基于上述研究,开发了动态数据驱动的河流突发性水污染事故预警系统技术平台,相关成果已集成到课题组研发的城市供水水质预警系统中,实现了动态数据驱动框架下的河流突发性水污染事故预警和应急监测点动态布局。该系统已在一个省和三个城市进行了部署和示范应用,已实现连续运行工作,为城市供水水质安全保障提供了信息支持。更多还原

【Abstract】 Water pollution accidents usually take place unexpectedly and may cause severe variations of raw water quality. To deal with this situation, the early warning information of water security is urgently needed by the departments of water supply. Especially, sudden water pollution accidents often occur abruptly with many uncertain factors. Thus, quick and precise prediction of the water variation trend is always a difficult and important issue in the research field of water environment. With the support of the National Science and Technology Major Project and the NSFC, this thesis mainly focuses on the methods and the system implementation techniques of early warning for sudden river pollution accidents.The main work and achievement of this thesis are as follows:(1) On the basis of Dynamic Data Driven Application System(DDDAS), a new framework of the early warning system for sudden river pollution accidents is proposed. The working principle of the newly framework is analyzed, and the differences compared with traditional ones in the application process are concluded. There are some attractive features and advantages in the new framework based on the DDDAS, such as multi-model integration, symbiosis feedback, dynamic nature, self learning function, which can be helpful to the water quality prediction and the decision making.(2) In order to solve the problems in the dynamic adaptability and convenience of the water quality prediction models, the dynamic early warning method of sudden river pollution accidents is studied under the proposed framework based on DDDAS. With the historical and dynamic monitoring data, the initial parameters of the prediction model are firstly adjusted using Case Based Reasoning (CBR) method and the initial early warning is realized. Then, in the case of more accurate prediction with inadequate hydrological and boundary data, using the improved grid search optimization method, a technique for dynamic optimization of parameters in the prediction model is implemented with the observed water quality data. As for the structural errors of the prediction model caused by the generalization process and other factors, the time series algorithm is used to realize the on-line correction of results of the prediction model. With the actual river pollution accidents as the application cases, the dynamic early warning method is demonstrated and analyzed. The application results show that the presented method is able to effectively supply early warning function for sudden river pollution accidents. Due to the dynamic correction mechanism, the deviation between the predicted results and the measured results can be reduced.(3) The current layout of emergency monitoring stations is usually lack of theory guidance, and the emergency departments are difficult to determine the location of the emergency monitoring stations fast and legitimately. To solve this problem, under the framework proposed based on the DDDAS, the dynamic method concerning the selection, evaluation and optimization of the emergency monitoring stations is studied. Using the Fuzzy Analytical Hierarchy Process (FAHP) algorithm, the technical solutions of the layout of emergent monitoring stations are evaluated dynamically. This method can supply the technical support for quantitative evaluation of the sensitivity and importance of emergent monitoring stations in different stages of sudden river pollution accidents. Thus, the limited emergency monitoring resources can be allocated more reasonably and timely. This technique is freshly researched in domestic and overseas.Based on the above research, the technique platform of dynamic data driven early warning system for sudden river pollution accidents is developed, which has been integrated into the urban drinking water quality early warning system developed by the research group. The early warning for sudden river pollution accidents and the dynamic layout of emergency monitoring stations based on DDDAS have been achieved. The urban drinking water quality early warning system has been deployed in one province and three cities. The systems are running well and have achieved continuous operation and have been deemed as information and technical supports for urban drinking water quality security.更多还原