Transferable surrogate models based on inductive biases of graph neural networks for water distribution systems

The occurrence of water losses in Water Distribution Systems is inevitable. Knowing that most of the real losses take place in distribution mains and in service connections, the methodology proposed in this paper is based on several leakage-assessment approaches from literature and on water distribution network modeling. This allows assessment of the benefits that can be achieved by pressure management in Water Distribution Systems, particularly in terms of water production reduction. Moreover, this approach can be useful for cost benefit analysis to help establish the level after which there is no more economic interest in reducing water losses (Economic Level of Leakage). Finally, the results from hypothetical case studies are presented and discussed, assuming the installation of Pressure Reducing Valves at District Metered Areas entry points.

Water losses in watersupply and distribution systems have been a serious problem in Turkey and allover the world causing economic costs and environmental drawbacks. Total waterlosses is the summation of real (physical) and apparent (commercial) losses. Mainactivities to reduce water losses include digitizing the components of network usingGeographical Information Systems (GIS), dividing large water distributionnetworks (WDNs) into District Metered Areas (DMAs), monitoring, carrying outyearly water balance, hydraulic modelling, pressure management, active leakagecontrol, recording water consumption of the users, reducing meter errors andillegal water consumption. The Ministry of Forestry and Water Affairs in Turkeyissued a new regulation for water losses reduction in 2014 based on the methodsused to determine and reduce the losses. This is followed by issuing practicaldescription of the technical procedure to reduce the losses in 2015. The newlegislation aims to reduce total water losses in all water authorities below 25%within certain time periods. All the water authorities are required to carryout yearly water balance and report them to the ministry together with forms tofollow up the efforts towards water losses reduction. In order to apply theprescribed techniques in a successful way, first of all water losses problemshould be understood well, technical and human resource capacity should beimproved, and sufficient financial resources should be allocated. Also, thewater utilities should have a strategic plan and a strong commitment to reducewater losses.

Node representation learning on attributed graphs-whose nodes are associated with rich attributes (e.g., texts and protein sequences)-plays a crucial role in many important downstream tasks. To encode the attributes and graph structures simultaneously, recent studies integrate pre-trained models with graph neural networks (GNNs), where pre-trained models serve as node encoders (NEs) to encode the attributes. As jointly training large NEs and GNNs on large-scale graphs suffers from severe scalability issues, many methods propose to train NEs and GNNs separately. Consequently, they do not take feature convolutions in GNNs into consideration in the training phase of NEs, leading to a significant learning bias relative to the joint training. To address this challenge, we propose an efficient label regularization technique, namely Label Deconvolution (LD), to alleviate the learning bias by a novel and highly scalable approximation to the inverse mapping of GNNs. The inverse mapping leads to an objective function that is equivalent to that by the joint training, while it can effectively incorporate GNNs in the training phase of NEs against the learning bias. More importantly, we show that LD converges to the optimal objective function values by the joint training under mild assumptions. Experiments demonstrate LD significantly outperforms state-of-the-art methods on Open Graph Benchmark datasets.

A realistic model extraction attack against graph neural networks

The aim of this study is to identify the most feasible pressure management (PM) method financially among four available PM methods [i.e., Fixed Outlet (FO), Time Modulated (TM), Flow Modulated (FM), and Remote Node Modulated (RNM) PM] recently introduced to EPANET source code by our previous study. For that, the four PM methods are tested in 18 different water distribution systems (WDSs), which differ in size, demand pattern, and amount of leakage. PM is inevitable for WDSs; if not applied, leakage flow rate, pipe bursts, active leakage control (ALC) efforts, and indirect water and energy losses increase abundantly. This circumstance increases expenditures due to augmentations in the amount of water entering the network and allocations from the municipal budget for pipe bursts, ALC activities, and energy consumption. Equipment and construction costs must be considered when PM is planned. However, leakage flows, pipe bursts, ALC activity efforts, and indirect water and energy losses are reduced with PM application, generating economic benefits. Four PM methods bring different levels of economic benefits. The cost-benefit analysis of each method has been performed, and the method that brings the maximum benefit (maximum cost reduction) has been determined. It has been revealed that the most applicable PM method, according to the cost-benefit analysis, varies according to the unit water cost. The RNM PM method is found to be the most appropriate when the unit water cost is high. As the unit water cost decreases, FO PM or TM PM becomes the most appropriate PM method instead of RNM PM. However, when the demand pattern of the network is smooth, the most feasible method becomes FO PM, and when the demand pattern is peaked while the unit water cost is relatively low, TM PM becomes the most applicable method.

Water distribution systems (WDS) face increasing challenges from climate change, urbanization, and growing populations, making efficient and accurate modeling crucial for their management. While traditional physics-based hydraulic models provide reliable results, they are computationally intensive, limiting their application in real-time decision-making and large-scale optimization. Our multi-year research journey demonstrates a progressive evolution in WDS modeling, successfully combining data-driven efficiency with fundamental physical principles.Our investigations evolve from conventional neural networks to increasingly sophisticated physically-informed approaches. We demonstrate how Graph Neural Networks can leverage network topology to improve prediction accuracy, but more importantly, how reformulating the problem in the edge space allows direct embedding of mass conservation principles. This novel Edge-Based Graph Neural Network (EGNN) architecture not only achieves superior performance but also demonstrates remarkable zero-shot generalization capabilities across different network configurations.Building on these insights, we further reformulate steady-state estimation as a diffusion process on graph edges, incorporating both mass and energy conservation laws. This physics-based reformulation enables direct GPU acceleration without relying on machine learning approximations, achieving near-perfect accuracy on multiple benchmarks while maintaining substantial computational speedups compared to traditional solvers.We then extend this approach to handle uncertainty by developing a topological Gaussian Process framework, where the covariance structure naturally encodes the physical conservation laws. This probabilistic extension enables rapid uncertainty quantification under variable demands, providing analytical uncertainty bounds without the computational burden of Monte Carlo sampling, while preserving the physical consistency guaranteed by our diffusion-based formulation.

This study proposes a method for the diagnosis of real water loss and for the actions which reduce this problem in water distribution systems. By adopting the parameters involved in the evaluation of real water loss, actions targeting combat and reduction of loss are prioritized, which allows for the optimization of the available resources. The topic of water loss reduction has recently been widely discussed, which is especially related to advances in information technology and to the availability of data within Brazilian water supply systems. The greater Sao Paulo area is one of the world’s most populated areas, with 20 million inhabitants. The Water System Cantatereira of Sao Paulo is very important in this area, supplying water to 60% of the entire population. In this system, one of the most relevant distributors is the Casa Verde Sector, composed of 84.2 kilometers of water distribution networks, 11,002 connections, and 16,891 homes, all with a water flow of 201.9 l/s. With the employment of geophones, listening stems and leak noise correlation, non-visible leaks were detected in this important and complex system, when water flow and pressure were recorded. All data was analyzed and correlated with information in the Geographic Information System (GIS). This study presents the results yielded by analyzes carried out with the software Water Audit 4. This software can be applied to any water supply system and is provided at no cost by the American Water Works Association (AWWA). It was concluded that the Casa Verde Sector must develop a specific strategy for water loss reduction, based on the renovation of the facilities and on the method of storage used for pressure data.

In this paper, we discuss the source of effectiveness of Graph Neural Networks (GNNs) in EDA, particularly in the VLSI design automation domain. We argue that the effectiveness comes from the fact that GNNs implicitly embed the prior knowledge and inductive biases associated with given VLSI tasks, which is one of the three approaches to make a learning algorithm physics-informed. These inductive biases are different to those common used in GNNs designed for other structured data, such as social networks and citation networks. We will illustrate this principle with several recent GNN examples in the VLSI domain, including predictive tasks such as switching activity prediction, timing prediction, parasitics prediction, layout symmetry prediction, as well as optimization tasks such as gate sizing and macro and cell transistor placement. We will also discuss the challenges of applications of GNN and the opportunity of applying self-supervised learning techniques with GNN for VLSI optimization.

Simultaneous optimization of energy and water quality in real-time large-sized water distribution systems is a daunting task for water suppliers. The complexity of energy optimization increases with a large number of pipes, scheduling of several pumps, and adjustments of tanks' water levels. Most of the simultaneous energy and water quality optimization approaches evaluate small (or hypothetical) networks or compromise water quality. In the proposed staged approach, Stage 1 uses a risk-based approach to optimally locate the chlorine boosters in a large distribution system based on residual chlorine failures and the associated consequences in different land uses of the service area. Integrating EPANET and CPLEX software, Stage 2 uses mixed integer goal programming for optimizing the day-ahead pump scheduling. The objective function minimizes the pumping energy cost as well as the undesirable deviations from goal constraints, such as expected water demand. Stage 3 evaluates the combined hydraulics and water quality performances at the network level. The implementation of the proposed approach on a real-time large-sized network of Al-Khobar City in Saudi Arabia, with 44 groundwater wells, 12 reservoirs, 2 storage tanks, 191 mains, 141 junctions, and 17 pumps, illustrated the practicality of the framework. Simulating the network with an optimal pumping schedule and chlorine boosters' locations shows a 40% improvement in water quality performance, desired hydraulics performance with optimal pump scheduling, and an average 20% energy cost reduction compared to the normal (unoptimized) base case scenario.

A surrogate evolutionary neural architecture search algorithm for graph neural networks

In many modern water networks, an emerging trend is to measure pressure at various points in the network for operational reasons. Because leaks typically induce a signature on pressure, these routine measurements can be used to develop nonintrusive leak detection approaches. This research employs successive linear approximation methods, based on linear programming and mixed integer linear programming, in a simulation-optimization framework to explore an alternative leak detection methodology for urban water distribution networks based on pressure measurements. The methods attempt to minimize the absolute differences between observed and simulated pressure values at the sensors to determine a linear combination of leaks that most closely approximates the observed pressure pattern. Steady-state and time-varying models of differing complexity (from small published networks to a 27,000-node network for a U.S. utility) were used to test the method. Results are presented to illustrate the method’s effectiveness under different conditions. The methods are shown to work well when pervasive pressure data and hydraulic models representing true operational conditions are available. The methods developed in this work are not intended to replace traditional leak detection methods; rather, they are meant to work in concert with available methods to more accurately and efficiently isolate leak locations and reduce water loss.

Introduction: Corrosion and scaling of water in distribution systems in spite of imposing economic and aesthetic problems can lead to adverse effects on consumers' health. The present study was conducted to evaluate the corrosion and scaling of water distribution system of Sarableh city (Ilam, Iran) during summer and winter seasons in 2014. Materials and methods: Totally, 60 samples of water (30 samples in each season) in 30 points of the distribution system were analyzed in terms of temperature, calcium hardness, alkalinity, TDS (total dissolved solids), pH, and DO (dissolved oxygen). Then, the corrosion and scaling of water were determined by various indices. Results: The results of the indices showed that the quality of drinking water in Sarableh city water distribution system had a medium corrosion rate. Conclusion: According to the results, it is needed to take some measures for stabilizing water before entering drinking water into the distribution system in order to prevent various health problems for consumers.

Online state estimation in water distribution systems via Extended Kalman Filtering

Water distribution systems are one of the critical infrastructures and major assets of the water utility in a nation. The infrastructure of the distribution systems consists of resources, treatment plants, reservoirs, distribution lines, and consumers. A sustainable water distribution network management has to take care of accessibility, quality, quantity, and reliability of water. As water is becoming a depleting resource for the coming decades, the regulation and accounting of the water in terms of the above four parameters is a critical task. There have been many efforts towards the establishment of a monitoring and controlling framework, capable of automating various stages of the water distribution processes. The current trending technologies such as Information and Communication Technologies (ICT), Internet of Things (IoT), and Artificial Intelligence (AI) have the potential to track this spatially varying network to collect, process, and analyze the water distribution network attributes and events. In this work, we investigate the role and scope of the IoT technologies in different stages of the water distribution systems. Our survey covers the state-of-the-art monitoring and control systems for the water distribution networks, and the status of IoT architectures for water distribution networks. We explore the existing water distribution systems, providing the necessary background information on the current status. This work also presents an IoT Architecture for Intelligent Water Networks - IoTA4IWNet, for real-time monitoring and control of water distribution networks. We believe that to build a robust water distribution network, these components need to be designed and implemented effectively.

High level water tank is a commonly used facility to satisfy water demand for people living in high building, which is also the so-called secondary water supply system (SWSS) in China. Due to lack of maintenance, water quality problem has been frequently reported in the SWSS. In the last few decades, anti-negative pressure facility (ANPF) has emerged as a revolutionary approach for solving the pollution in the SWSS. ANPF makes good use of the municipal pipeline pressure, which is the improvement of traditional water supply. However, most of recent research activities mainly focus on the influence of part of water distribution system (WDS). In this paper, the software of EPANET has been applied in the SWSS calculation and simulation. Besides that, the SWSS with ANPF simulation method is proposed to evaluate the effect of other all nodes pressure in the WDS using hydraulic model. As the cases study, the paper analyzed two WDSs with ANPF. After simulation of the SWSS, the result of WDS, which is smaller, indicates that 6 out of the 9 nodes are affected with ANPF and that of WDS, which is larger, demonstrates that 37 out of the 41 nodes are affected and the affected nodes are focused on the confluent of pipes, especially around the ANPF.