【作者】 王婷婷；

【导师】 高金良；

【作者基本信息】 哈尔滨工业大学 ， 建筑与土木工程（专业学位）， 2017， 硕士

【摘要】 随着智能水表技术的发展,实时监控系统使供水部门可以拥有大量关于供水管网属性的数据。智能水表包括记录水量和通信系统两部分,可以实时传输和储存用水量数据。智能水表已经被广泛应用,大多数城市都具有这样的设备,然而智能水表传送的关于管网属性的数据,水司仅仅用于日常调度和经济效益考评,之后这些大量数据会被储存一段时间。智能水表连续传送数据,随着数量逐渐增大,水司会因为内存原因而把这些数据删除,同时删除的还有这些数据所含有的非常有价值的管网信息。随着数据挖掘技术的发展,我们有技术有能力处理分析这些数据,最大程度地挖掘数据所包含信息。分析这些数据有助于供水管网革新供水管网管理、计划和用户服务,更加充分利用水资源,保护水资源。本文根据DMA分区流量数据特点,提出一种聚类方法,即基于DMA分区用水量曲线距离和形状的聚类算法(KS),该聚类方法相对经典K-means、自主映射(SOM)和模糊C均值而言,更能体现DMA分区用水量规律。通过Y市DMA分区项目中获得43个DMA分区的流量数据,对这43个DMA分区流量数据进行数据预处理之后,进行聚类分析,比较KS、K-means、SOM和FCM四种聚类算法效果,最终表明KS的聚类效果最好,并且通过分析KS聚类结果,能够指导水司检测异常情况(漏损、偷水)。在对43个DMA分区流量数据处理过程中,通过观察43个DMA分区的用水量变化曲线,发现根据《给水工程》等教材计算出的时变化系数,小于大多数各小时用水量占全天总用水量比例。说明若继续采用《给水工程》等教材中的时变化系数公式,将不能保证如Y市这样城市的供水安全,建议进一步修正时变化系数公式。更多还原

【Abstract】 The increasing use of smart water metering technologies for monitoring networks in real-time is providing water utilities with an ever growing amount of data on their business operations and infrastructure.Such metering devices embrace two distinct technologies: meters that record water usage.Although now mostly cities already have such smart meters,but the collected traffic data is only conducive for water supply network daily scheduling and water company economic benefit evaluation.These large data is stored to the database.But because of memory and so on,these large data is delete and the valuable information in data disappear after a period of time.This traditional approach has failed to keep pace with the current data age and cause resources to be wasted.With the development of data mining technology,we will lose opportunity to realize water supply network better,this problem is desperately needs to be solved and we should pay more attention on this problem.Analyzing these data will help the water supply network to innovate water supply network management,planning and customer service,make the most of water and protect water resource.This paper is based on the characteristics of DMA partition flow data,a clustering method based on the DMA water partition curve distance and shape(KS),The clustering method is relatively classic K – means,autonomous mapping(SOM)and fuzzy c-means,more can reflect the DMA partition law of water consumption.43 DMA traffic data come form DMA partition of Y city.The 43 DMA partition traffic data after data preprocessing,clustering analysis.Compare Clustering algorithm effect of KS,K-means,SOM and FCM.Finally,KS has the best clustering effect,By analyzing KS clustering results,that can guide water to detect anomalies(leakage,steal water).In the processing of 43 DMA partition flow data,observe the water consumption curve of 43 DMA partition,Found according to the water supply engineering and other teaching materials to calculate the variation coefficient that is lesser than most of the water every hour of total water consumption throughout the day.If continue to use the teaching material such as the water supply engineering of the variation coefficient formula,will not be able to ensure that cities like Y city water supply security,suggest that change correction coefficient formula.更多还原