Abstract Water leakage in water distribution networks (WDNs) can cause serious economic and social impacts. Therefore, it is urgent to discover the efficient sensor deployment methods to monitor the leakage of WDNs. While related research studies mainly consider the single-point leakage, this paper firstly presents a pressure sensor deployment strategy to deal with the problem of the multi-point leakage occurred in WDNs. The novel proposed label space dimension reduction joint mutual information method (LSDR-JMI) is based on a filter multi-label feature selection algorithm. The label space dimension reduction (LSDR) method is based on the dependence of the label matrix on the leak cases to obtain the optimal label subspace. Then the optimal subset of feature matrix is calculated according to the relevant information between the optimal label subspace and the feature matrix. This method is applied to two hydraulic networks simulated by EPANET to demonstrate its effectiveness and advantage.

中文翻译：

---

基于LSDR-JMI方法的配水管网新型多泄漏传感器部署策略

摘要 供水网络 (WDN) 中的漏水会造成严重的经济和社会影响。因此，迫切需要发现有效的传感器部署方法来监测 WDN 的泄漏。在相关研究主要考虑单点泄漏的同时，本文首先提出了一种压力传感器部署策略来处理WDN中发生的多点泄漏问题。新提出的标签空间降维联合互信息方法（LSDR-JMI）基于过滤器多标签特征选择算法。标签空间降维（LSDR）方法是基于标签矩阵对泄漏情况的依赖来获得最优标签子空间。然后根据最优标签子空间与特征矩阵之间的相关信息计算特征矩阵的最优子集。将该方法应用于EPANET模拟的两个水力网络，证明其有效性和优势。