Search

Currently, in the water distribution systems literature, fault detection methods are typically evaluated on benchmark water networks that do not include real-time experimental data, or on private commercial datasets, which prohibit the reproducibility of the results. Moreover, realistic modeling of faults on hydraulic system components, sensors and actuators is often unavailable. In this work, we provide a framework for the application of fault-diagnosis methodologies on WaterSafe, a water network benchmark for fault diagnosis. The WaterSafe benchmark is a small scale replica of a water transport network constructed using industrial components and devices, while the communications are implemented in a way that resemble a water utility's Supervisory Control and Data Acquisition system. A general problem formulation for fault-diagnosis on water systems is provided, in accordance to the mathematical model of the benchmark. Moreover, we provide a calibrated simulation model including system, sensor and actuator faults, based on observations from the real system. Finally, we provide open access to the datasets generated from the experiments containing the aforementioned faults. &nbsp

Includes bibliography (p. 143-148). Number of sources in the bibliography: 200 Thesis (Ph. D.) -- University of Cyprus, Faculty of Engineering, Department of Electrical and Computer Engineering, 2020. The University of Cyprus Library holds the printed form of the thesis. Οι σύγχρονες κοινωνίες αρχίζουν να συνειδητοποιούν τη σημασία του πόσιμου νερού λόγω της λειψυδρίας σε πολλά μέρη του κόσμου ως αποτέλεσμα του αυξανόμενου ανθρώπινου πληθυσμού και της κλιματικής αλλαγής. Η βιομηχανία ύδρευσης εκσυγχρονίζεται με την εγκατάσταση αισθητήρων για την παρακολούθηση και τον έλεγχο των ΣΔΝ και υπολογιστικών συστημάτων για την επεξεργασία αυτών των δεδομένων. Οι αλγόριθμοι εκτίμησης κατάστασης συμπεραίνουν την κατάσταση του συστήματος, όπως τις ροές νερού και τις πιέσεις στους αγωγούς, τα επίπεδα των υδατοδεξαμενών και την συγκέντρωση χλωρίου στο νερό, χρησιμοποιώντας τις διαθέσιμες μετρήσεις και μοντέλα του συστήματος. Η πλήρης εκτίμηση της κατάστασης του δικτύου διανομής σε πραγματικό χρόνο υποστηρίζει τη διαδικασία λήψης αποφάσεων και καθιστά δυνατή την αποτελεσματική λειτουργία αυτών των συστημάτων, βελτιώνει την εξυπηρέτηση των πελατών και καθιστά δυνατή την έγκαιρη ανίχνευση και ανταπόκριση σε επείγουσες καταστάσεις όπως σε σπασίματα σωλήνων ή μόλυνση του νερού, ελαχιστοποιώντας έτσι τις επιπτώσεις τους. Ωστόσο, η εκτίμηση της κατάστασης στα ΣΔΝ είναι ένα δύσκολο έργο, κυρίως λόγω της έλλειψης μετρήσεων, που καθιστά τα μεγάλης κλίμακας συστήματα ύδρευσης μη παρατηρήσιμα. Η ποιότητα του νερού είναι εγγενώς συνδεδεμένη με την υδραυλική δυναμική, γεγονός που καθιστά τα μοντέλα ποιότητας νερού χρονομεταβλητά. Επιπλέον, η αβεβαιότητα στα μοντέλα και στις παραμέτρους ποιότητας νερού δυσκολεύουν περαιτέρω τον ακριβή χαρακτηρισμό της ποιότητας στα ΣΔΝ. Αυτή η διατριβή παρουσιάζει μεθοδολογίες για την εκτίμηση της κατάστασης των ΣΔΝ, οι οποίες απευθύνουν τις παραπάνω πρακτικές προκλήσεις. Οι προτεινόμενες μεθοδολογίες χρησιμοποιούν διαθέσιμες πληροφορίες για τα συστήματα αυτά αποσκοπώντας στην εξαγωγή αξιόπιστων εκτιμήσεων κατάστασης, οι οποίες μπορούν να χρησιμοποιηθούν από εταιρείες παροχής νερού και επαγγελματίες. Συγκεκριμένα, προτείνεται μια μέθοδος εκτίμησης υδραυλικής κατάστασης η οποία προϋποθέτει οριοθετημένες αβεβαιότητες στις μετρήσεις και στα μοντέλα και εκτιμά ένα εύρος τιμών για την κατάσταση του συστήματος λαμβάνοντας υπόψη τη συνδυασμένη επίδρασή τους. Η διαθεσιμότητα εκτιμήσεων υδραυλικής κατάστασης επιτρέπει επιπλέον την εκτίμηση της ποιότητας του νερού, λόγω της διασυνδεδεμένης φύσης της υδραυλικής δυναμικής και της δυναμικής ποιότητας του νερού. Παρέχεται μια μέθοδος εκτίμησης κατάστασης της ποιότητας του νερού η οποία υπολογίζει τα όρια της συγκέντρωσης χλωρίου και είναι σε θέση να ενσωματώσει αλγόριθμους για εκτίμηση των ποιοτικών παραμέτρων σε πραγματικό χρόνο, λαμβάνοντας υπόψη την ύπαρξη άγνωστων παραμέτρων ποιότητας νερού. Ο κύριος στόχος των προτεινόμενων πρακτικών προσεγγίσεων εκτίμησης κατάστασης είναι να καταστεί δυνατή η ανάπτυξη μεθοδολογιών για τη διάγνωση σφαλμάτων στα ΣΔΝ. Συγκεκριμένα, στοχεύουμε δύο τύπους κρίσιμων σφαλμάτων στα ΣΔΝ: τις διαρροές νερού και την μόλυνση του νερού. Προτείνονται δύο μεθοδολογίες για την ανίχνευση και τον εντοπισμό διαρροών με βάση την εκτίμηση εύρους τιμών για την υδραυλική κατάσταση του συστήματος και αξιολογούνται χρησιμοποιώντας ένα δείκτη αναφοράς διαρροής ανοικτής πρόσβασης και καθορισμένα δίκτυα αναφοράς. Η ανίχνευση μολύνσεων στο νερό επιτυγχάνεται συνδυάζοντας τις προτεινόμενες προσεγγίσεις εκτιμήσεως κατάστασης των υδραυλικών και ποιότητας νερού με μια λογική ανίχνευσης που χρησιμοποιεί κατώφλια πολλαπλών επιπέδων. Επιπλέον, διερευνάμε την έννοια της ενεργής ανίχνευσης βλαβών με την ανάπτυξη μιας νέας μεθοδολογίας για την επέκταση της περιοχής που παρακολουθείται από αισθητήρες ποιότητας νερού, η οποία αντιμετωπίζει το πρόβλημα της έλλειψης μετρήσεων. Modern societies are growing more conscious of the importance of drinking water, due to water scarcity in many parts of the world as an effect of an increasing human population and climate change. As a result of aging infrastructures, water utilities face problems which were ignored in the past such as high rates of water loss and insufficient monitoring of water quality. The water industry is being modernized with the installation of sensors for monitoring and control of WDS and computer systems to process these data. State estimation algorithms infer the state of the system, such as water flows and pressures in pipes, tank levels and chlorine concentration in water, using the available measurement-set and system model. A complete view of the distribution network state in real-time supports the decision-making process and enables the efficient operation of these systems, improves customer service and enables the early detection and response to emergency events such as pipe failures or water contamination, thus minimizing their impact. However, state estimation in WDS is a challenging task, mainly due to measurement scarcity which makes these large-scale systems unobservable. Water-quality is intrinsically connected with hydraulic dynamics, making water-quality models time-varying. Additionally, modeling and parameter uncertainty further increase the difficulty in accurately characterizing water-quality in WDS. This thesis presents methodologies for state estimation in WDS that address the above practical challenges. The proposed methodologies use available a priori information about these systems to derive reliable state estimates which can be utilized by water utilities and practitioners. Specifically, a hydraulic-state estimation approach is proposed which assumes bounded measurement and modeling uncertainties and calculates an interval-state estimate considering their combined effect. The reasoning behind this approach is that providing a range of possible values for each state depending on the level of modeling uncertainty, is often more useful to an operator than providing estimates with no information about the estimation error. The availability of hydraulic-state estimates enables water-quality estimation, due to the interconnected nature of hydraulic and water-quality dynamics. A water-quality state estimation approach is presented which calculates bounds on chlorine concentration and is able to incorporate real-time parameter estimation, considering the existence of unknown water-quality parameters. The aim of the proposed state estimation approaches is to obtain reliable bounds on state estimates, in real-time, which is a key component in many methodologies related to model-based fault diagnosis. Specifically, two type of critical faults in WDS are targeted in this thesis: water leakages and water contamination. Two methodologies for detecting and localizing leakages based on hydraulic interval-state estimation are proposed and evaluated using an open access leakage diagnosis benchmark and established benchmark networks. Water contamination detection is achieved by combining the proposed hydraulic and water-quality state estimation approaches with a detection logic using multi-level thresholds. Moreover, we explore the concept of active fault detection by developing a novel methodology for enhancing the area monitored by water-quality sensors, to address the problem of measurement scarcity.

State estimation of hydraulics (i.e. pressure and flows) in water distribution networks is an important tool for efficient and resilient operation.However, hydraulic state estimation is a challenging task in practice due to the scarcity of measurements and the presence of several modeling uncertainties.Standard state estimation techniques may produce unreliable estimates with no information of the estimation error magnitude, especially when historical data are used in place of missing measurements.In this paper, we propose a comprehensive methodology for generating hydraulic state bounding estimates by considering both measurement and parametric uncertainties.The methodology is based on solving the nonlinear interval hydraulic equations using bounding linearization, a technique that restricts the nonlinearities within a convex set, thus converting the problem in a form which is solvable using linear optimization. An iterative procedure improves the bounding linearization, converging to the tightest possible bounds.Simulation results demonstrate that the proposed methodology produces tight state bounds that can replace Monte-Carlo simulations.

Hydraulic state estimation in water distribution networks is the task of estimating water flows and pressures in the pipes and nodes of the network based on some sensor measurements. This requires a model of the network as well as knowledge of demand outflow and tank water levels. Due to modeling and measurement uncertainty, standard state estimation may result in inaccurate hydraulic estimates without any measure of the estimation error. This paper describes a methodology for generating hydraulic state bounding estimates based on interval bounds on the parametric and measurement uncertainties. The estimation error bounds provided by this method can be applied to determine the existence of unaccounted-for water in water distribution networks. As a case study, the method is applied to a modified transport network in Cyprus, using actual data in real time.

Contamination event detection in Water Distribution Systems (WDS) is an important part of the overall procedures followed by operators to ensure the delivery of safe drinking water to consumers. In practice, detection methodologies may use a small number of sensors, which may be placed optimally throughout the network, to monitor water quality. Typically, due to their high costs, only a small number of these sensors are available within each WDS, and as a result, a part of the network is not monitored by sensors. In this paper, we propose a novel methodology for detecting contamination events by increasing the area monitored by these water quality sensors. Specifically, the proposed methodology can be used in emergency cases when information is available about a possible contamination event in the system, to actively manipulate WDS actuators, by closing and opening valves, and alter the flow directions within the network. This is based on the concept of Active Fault Detection, in which the control input of a system is modified with the aim to improve detectability of a fault. This active detection scheme, drives flows from specific parts of the network in pre-determined paths, to allow the sensors to monitor the quality of water from previously unobserved parts of the network. As a result, the monitoring coverage of the sensors is increased and some contamination events occurring within those areas can be detected. Moreover, the methodology facilitates the isolation of the contamination propagation path and its possible source. We demonstrate how such goals can be achieved on two simple example networks, discuss the benefits of the results and open the discussion for further work in this area.

This work develops a new algorithm based on interval hydraulic state estimation for leakage detection in urban water distribution systems using pressure sensors and smart meters. The proposed approach takes into consideration the fact that water demand and network parameter uncertainties are unstructured, and models them as intervals defined by a corresponding upper and lower bound. The state estimation process yields bounded state estimates which are used in the creation of detection thresholds. A detection logic is used to determine whether the thresholds are violated as a result of a leakage fault event. The proposed methodology is demonstrated on a leakage benchmark dataset under various scenarios., © 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. \\ S. G. Vrachimis, D. G. Eliades and M. M. Polycarpou, "Leak Detection in Water Distribution Systems Using Hydraulic Interval State Estimation," 2018 IEEE Conference on Control Technology and Applications (CCTA), Copenhagen, 2018, pp. 565-570. doi: 10.1109/CCTA.2018.8511516

Water distribution networks are critical infrastructure systems, which are required to reliably deliver water to consumers. Hydraulic state estimation of flows and pressures is key for the efficient and dependable operation of water systems. However, in practice, hydraulic state estimation is a challenging task due to the scarcity of sensors compared to system states and the presence of several modeling uncertainties. It is a common practice for historical data to be used in the place of missing measurements, however the lack of a statistical characterization for the error of these measurements results in having no knowledge of the state estimate error magnitude. This paper presents a methodology for generating hydraulic state bounding estimates by modeling measurement and parametric uncertainties as intervals. The resulting nonlinear interval hydraulic equations are re-formulated using bounding linearization, a technique that restricts the nonlinearities within a convex set, and solved using linear optimization. An iterative procedure improves the bounding linearization, converging to the tightest possible bounds. Simulation results demonstrate that the proposed methodology produces tight state bounds and can be successfully used to detect water leakages in the network.

Contamination event detection in Water Distribution Systems (WDS) is an important part of the overall procedures followed by operators to ensure the delivery of safe drinking water to consumers. In practice, detection methodologies may use a small number of sensors, which may be placed optimally throughout the network, to monitor water quality. Typically, due to their high costs, only a small number of these sensors are available within each WDS, and as a result, a part of the network is not monitored by sensors. In this paper, we propose a novel methodology for detecting contamination events by increasing the area monitored by these water quality sensors. Specifically, the proposed methodology can be used in emergency cases when information is available about a possible contamination event in the system, to actively manipulate WDS actuators, by closing and opening valves, and alter the flow directions within the network. This is based on the concept of Active Fault Detection, in which the control input of a system is modified with the aim to improve detectability of a fault. This active detection scheme, drives flows from specific parts of the network in pre-determined paths, to allow the sensors to monitor the quality of water from previously unobserved parts of the network. As a result, the monitoring coverage of the sensors is increased and some contamination events occurring within those areas can be detected. Moreover, the methodology facilitates the isolation of the contamination propagation path and its possible source. We demonstrate how such goals can be achieved on two simple example networks, discuss the benefits of the results and open the discussion for further work in this area.

Hydraulic state estimation in water distribution networks is the task of estimating water flows and pressures in the pipes and nodes of the network based on some sensor measurements. This requires a model of the network, as well as knowledge of demand outflow and tank water levels. Due to modeling and measurement uncertainty, standard state-estimation may result in inaccurate hydraulic estimates without any measure of the estimation error. This paper describes a methodology 5 for generating hydraulic state bounding estimates based on interval bounds on the parametric and measurement uncertainties. The estimation error bounds provided by this method can be applied to determine the existence of unaccounted-for water in water distribution networks. As a case study, the method is applied to a modified transport network in Cyprus, using actual data in real-time., Vrachimis, S. G., Eliades, D. G., and Polycarpou, M. M.: Real-time hydraulic interval state estimation for water transport networks: a case study, Drink. Water Eng. Sci., 11, 19-24, https://doi.org/10.5194/dwes-11-19-2018, 2018.

Contamination event detection in Water Distribution Systems (WDS) is an important part of the overall procedures followed by operators to ensure the delivery of safe drinking water to consumers. In practice, detection methodologies may usea small number of sensors,whichmaybeplaced optimally throughout the network, to monitor water quality. Typically, due to their high costs, only a small number of these sensors are available within each WDS, and as a result, a part of the network is not monitored by sensors. In this paper, we propose a novel methodology for detecting contamination events by increasing the area monitored by these water quality sensors. Specifically, the proposed methodology can be used in emergency cases when information is available about a possible contamination event in the system, to actively manipulate WDS actuators, by closing and opening valves, and alter the flow directions within the network. This is based on the concept of Active Fault Detection, in which the control input of a system is modified with the aim to improve detectability of a fault. This active detection scheme, drives flows from specific parts of the network in pre-determined paths, to allow the sensors to monitor the quality of water from previously unobserved parts of the network. As a result, the monitoring coverage of the sensors is increased and some contamination events occurring within those areas can be detected. Moreover, the methodology facilitates the isolation of the contamination propagation path and its possible source. We demonstrate how such goals can be achieved on two simple example networks, discuss the benefits of the results and open the discussion for further work in this area.