Publication: A PROBABILISTIC FRAMEWORK FOR HYDRAULIC PERFORMANCE ASSESSMENT OF COMPLEX WATER DISTRIBUTION NETWORKS
Date
2019
Authors
JEREZ URQUIETA, DANKO JOSÉ
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
En este trabajo se formula e implementa un marco probabilístico para la evaluación del desempeño hidráulico de redes complejas de distribución de agua. Se consideran dos tipos de problemas en este contexto: evaluación de la con fiabilidad hidráulica, y problemas relacionados a detección. Por un lado, se desarrolla e implementa un enfoque estocástico parallevar a cabo la evaluación de con abilidad hidráulica de sistemas complejos de distribuciónde agua. La con abilidad hidráulica es una medida probabilística del nivel de desempeñohidráulico de un sistema de distribución de agua, que involucra problemas altamente dimensionalesdesde el punto de vista probabilístico. La efectividad del método propuestoes evaluado mediante el estudio de una red de distribución real con miles de nodos y tuberías. Se demuestran varios tipos de análisis posibles en el marco propuesto, incluyendopropagación de incertidumbre, análisis de confi abilidad y sensibilidad, y análisis de falla.Por otro lado, se desarrolla una metodología de actualización Bayesiana de modelos basadaen simulación para enfrentar problemas relacionados a detección, tales como el localizarfugas. Este tipo de problemas se formula como un problema de selección de clase de modeloprobabilístico, en donde un cierto número de clases de modelos se de nen como potencialescaracterizaciones de la condición real de la red. La clase más probable se determina enbase a las mediciones disponibles y el conocimiento previo sobre el sistema. Dicha clasese interpreta como la condición real de la red. Con el fi n de demostrar la aplicabilidad delenfoque propuesto, se formulan y estudian dos ejemplos de aplicación. Estos correspondena los problemas de detección de fugas y detección de conectividad, involucrando redesreales de distribución de agua y mediciones sintéticas. Se estudia la incidencia de variosfactores en la efectividad de la metodología propuesta, incluyendo errores de medición ymodelación, así como la con guración de los dispositivos de monitoreo. Con el fi n de formulare implementar las metodologías propuestas, dos técnicas de simulación estocásticase han adaptado apropiadamente e integrado con un software hidráulico de nivel comercial.Los resultados obtenidos demuestran que el marco probabilístico desarrollado en estetrabajo provee herramientas e cientes y efectivas que permiten un entendimiento valiosodel desempeño hidráulico de redes complejas de distribución de agua.
A probabilistic framework for evaluating the hydraulic performance of complex waterdistribution networks is formulated and implemented in this work. Two classes of problemsare addressed: hydraulic reliability assessment, and detection-related problems. On onehand, a stochastic approach is developed and implemented in order to handle the hydraulicreliability assessment of large-scale water distribution systems. Hydraulic reliability is aprobabilistic measure of the hydraulic performance of a water distribution system, involvinghigh-dimensional problems from the probabilistic point of view. The eectiveness ofthe approach is evaluated with the study of a real-life water distribution network consistingof thousands of nodes and pipes. A number of analyses within the proposed frameworkare demonstrated, including reliability and sensitivity analysis, uncertainty propagationand failure analysis. On the other hand, a simulation-based Bayesian model updatingmethodology is developed to deal with detection-related problems in complex water utilitynetworks, such as leakage detection. Detection-related problems are expressed as modelclass selection problems, where a number of model classes are de ned as potential characterizationsof the actual network condition. Revised information based on available datais then used to select the most probable model class, which is identi ed as the actualnetwork condition. Two application examples are studied in order to evaluate the capabilitiesof the proposed approach. These examples correspond to leakage detection andconnectivity detection problems, involving real-life water distribution networks and syntheticmeasurement data. Several factors aecting the eectiveness of the approach areevaluated, including modeling uncertainties, measurement errors and sensors con guration.In order to set and formulate the proposed approaches, advanced simulation techniquesare properly adapted and integrated with a commercial level hydraulic simulator. Resultsdemonstrate that the framework provides functional tools that give a valuable insight intothe hydraulic performance of a class of complex utility networks.
A probabilistic framework for evaluating the hydraulic performance of complex waterdistribution networks is formulated and implemented in this work. Two classes of problemsare addressed: hydraulic reliability assessment, and detection-related problems. On onehand, a stochastic approach is developed and implemented in order to handle the hydraulicreliability assessment of large-scale water distribution systems. Hydraulic reliability is aprobabilistic measure of the hydraulic performance of a water distribution system, involvinghigh-dimensional problems from the probabilistic point of view. The eectiveness ofthe approach is evaluated with the study of a real-life water distribution network consistingof thousands of nodes and pipes. A number of analyses within the proposed frameworkare demonstrated, including reliability and sensitivity analysis, uncertainty propagationand failure analysis. On the other hand, a simulation-based Bayesian model updatingmethodology is developed to deal with detection-related problems in complex water utilitynetworks, such as leakage detection. Detection-related problems are expressed as modelclass selection problems, where a number of model classes are de ned as potential characterizationsof the actual network condition. Revised information based on available datais then used to select the most probable model class, which is identi ed as the actualnetwork condition. Two application examples are studied in order to evaluate the capabilitiesof the proposed approach. These examples correspond to leakage detection andconnectivity detection problems, involving real-life water distribution networks and syntheticmeasurement data. Several factors aecting the eectiveness of the approach areevaluated, including modeling uncertainties, measurement errors and sensors con guration.In order to set and formulate the proposed approaches, advanced simulation techniquesare properly adapted and integrated with a commercial level hydraulic simulator. Resultsdemonstrate that the framework provides functional tools that give a valuable insight intothe hydraulic performance of a class of complex utility networks.
Description
Catalogado desde la version PDF de la tesis.
Keywords
MARCO PROBABILISTICO , REDES COMPLEJAS , TECNICAS DE SIMULACION AVANZADA