基于复杂网络聚类的给水管网高维多目标分区优化研究

District metering areas (DMAs) is an effective technology to control water leakage in water distribution systems (WDSs). There are many factors to be considered when designing DMAs. Traditional methods consider very few factors and mainly use engineering principals and hydraulic simulation to validate feasibility, without considering the structure, therefor it can not achieve optimal DMAs. Optimization is an alternative method. However, there are very few investigations in optimal DMAs design, the main reasons are due to the complex structure of WDSs and many-objective evaluation in DMAs design. We have explored network structure for DMAs with engineering principles before. In this research applicability of many-objective DMAs design will be investigated deeply based on the inner structure of WDSs, emphases are placed on complex network clustering and many-objective optimization design for DMAs. First, the transmission pipes are identified and topology matrix of WDSs are simplified by shortest path algorithm. Based on complex network clustering, candidate DMAs are obtained using many-objective optimization with valve and meter location as two decisions, hydraulic requirement and water quality are met simultaneously. At last global and local pressure management are combined to assess the rationality of candidate DMAs with economic analysis of pressure management. It is anticipated that the research will provide some suggestions for investigation of WDSs structure and water loss control by DMAs optimization design.

给水管网分区是控制漏损的有效技术措施，涉及众多影响因素。传统方法依据工程经验并通过水力模拟验证分区可行性，涉及目标少且未考虑系统结构内在联系，因此无法实现最优分区。而优化方法为科学分区提供了新的途径。目前针对优化分区的研究较少，主要原因是给水管网的复杂结构以及分区评价的高维多目标性。我们已经结合分区原则从管网结构上探索分区优化设计。本项目拟从给水系统结构内部联系上进一步研究高维多目标分区的适用性，着重实现符合给水管网特性的复杂网络聚类分析和高维多目标优化分区方法。首先通过最短路径算法确定输水干管、简化给水管网拓扑矩阵，然后以复杂网络聚类为基础，以阀门和水表位置为决策变量，采用高维多目标优化算法，初步确定满足水力约束和水质约束的候选分区方案。继而以压力管理经济性为目标，通过全局压力管理与区域精准压力控制相结合的综合效果评价候选方案的合理性。为给水系统结构研究和分区优化漏损控制提供一些建议。

许多供水管网都有严重的漏损问题，这使得给水管网分区在漏损控制中的作用日益受到重视。目前分区方法的有效性难以满足实际给水管网漏损管理的需求，如何快速有效地实现管网自动分区成为本研究领域亟待攻克的难点问题。本项目针对给水管网自动分区问题，实现了给水管网分区的理论方法，建立了分区多标准评价、高维多目标优化和分区后性能评价等多种评价指标体系。1）通过给水管网拓扑特征提取，提出了基于谱聚类和遗传算法的分区节点聚类，基于给水管网节点和管道中心性分析，优化确定了分区水表和边界阀门的位置，并通过模块度、分区均匀性、分区水压和水头均匀性、可靠性等指标对优化分区进行多标准评价，确定合理的分区。2）提出管道水力指数综合评价管道水力特性，并以此指数为权重，采用最小生成树确定了管网主要枝状供水路径，在此基础上，分别优化确立了主、次进水管道和边界阀门位置，最后以系统弹性和水龄为指标，对比分析分区方案。3）针对分区后性能变化问题，研究了分区在漏损、消防流量和水质等方面的性能。本项目研究成果的意义在于，建立了复杂给水管网分区的理论方法和分区性能评价指标体系。有效克服了传统人工分区方法的局限，从理论上有助于进一步推动管网分区向着更高效、更智能的方向发展，从应用上实现管网分区在漏损控制、水质改善、提升弹性和可靠性等方面的进步，实现供水管网的优化运行，提高供水系统效率，保障供水系统安全。