João Carlos Rosa Marques ROBUST DESIGN OF WATER DISTRIBUTION NETWORKS FOR A PROACTIVE RISK AND UNCERTAINTY MANAGEMENT Doctoral Thesis in Civil Engineering, with specialization in Hydraulics, Water Resources and Environment, submitted to the Faculty of Sciences and Technology of the University of Coimbra and supervised by Professor Doctor Maria da Conceição Morais de Oliveira Cunha and Professor Doctor Dragan Savić October,2013 Financial support This research work was financed by “Fundação para a Ciência e a Tecnologia” (FCT, Portugal) through the PhD grant with reference SFRH / BD / 47602 / 2008, and was co- financed by the European Social Fund (ESF) within the program “Programa Operacional Potencial Humano (POPH). The POPH research program is integrated in the “National Strategic Reference Framework 2007-2013 - Tipologia 4.1 – Formação Avançada” (QREN 2007-2013). Acknowledgments I want to acknowledge my supervisors, Professor Maria da Conceição Cunha for having challenged me to do my PhD. Her permanent support since the beginning was essential to guide me during this journey. PhD is a complex process and her constant encouragements and work ethics helped me not only on a professional level but also in growing as a person. Also, I want to acknowledge Professor Dragan Savić for his openness’s and incentives, for his hospitality during my stay in Exeter and for sharing with me several interesting publications and his vast knowledge. My gratitude goes to Fundação para a Ciência e Tecnologia (FCT) who provided me financial support throughout my study. To the University of Coimbra and to IMAR I want to express my thankfulness for the institutional support and for providing me a place to work. I would like to thank to all my colleagues at the Department of Civil Engineering. During these years we shared not only a working place but also very good discussions and moments. I’m really grateful to my friend and colleague of the same research project, Ina Vertommen, for her editing work to improve this thesis. A special mention should also go to Professor Joaquim Sousa for his insights and discussions about hydraulic simulators and some specific case studies. I also want to give thanks to many colleagues in the Centre of Water Systems in Exeter, who enriched my stay there, especially to Andrew Duncan who showed me some fabulous places on the South West of England. To my family, for all the support and their permanent encouragement, I want to express my thankfulness. Finally I want to deeply thank my wife, and mother of my two most precious treasures, not only for her support given to me during this journey, but also for all the moments and dreams we share together. v Abstract Water supply systems are crucial infrastructures in every modern society. In the last three decades the optimal design of these systems has been studied by a great number of academics and water industry professionals. However, most of these studies, consider future as predefined and working conditions perfectly known. This simplification leads to solutions which although optimal for the imposed conditions, may perform badly if reality turns out to be considerably different. In real world the working conditions can be disturbed by different types of situations such as, pipe breaks or tank failures, energy supply interruptions or changes in demand. Only a proactive attitude towards risk and uncertainty can deal with these aspects from the design phase until the end of the life cycle of the different components. Decision makers have to decide how these infrastructures should be planned and operated to cope with uncertain future situations. Researchers can provide help in finding and justifying good solutions, through the investigation of optimization tools. The main contributions of this work are the decision-aid tools developed for the optimization of design and operation of water distribution networks capable of dealing with risk and uncertainty, thus being able to find more robust and reliable solutions. Water systems are costly and it is difficult to change physical component conditions without making large investments. When planning is considers future uncertainty, it is possible to obtain better solutions for a long planning horizon. In the initial chapters, this work provides some insight into concepts and definitions regarding water distribution systems, as well as a comprehensive state of the art on vii optimization techniques and on decision support approaches. Two optimization tools are also presented: the single-objective and multi-objective simulated annealing optimization methods. Decision-aid tools are then proposed. A robust optimization model for the optimal design of water supply systems operating under different circumstances is shown first. The results show that robustness can be included in several ways and for varying levels of reliability. The inclusion of robustness considerations leads to more reliable designs, while at the same time results in small cost increases. This is followed by a group of three different decision-aid models based on a real options approach, which constitute the main scientific contribution of this thesis. The first optimization model based on real options is used for the optimal design and operation of water distribution networks considering various possible future uncertainties. The results of this its implementation on a water network show that building flexibility into the decision strategy enables an adaptive approach to be taken that can avoid future problems with network capacity. The second optimization model is proposed to optimize water distribution networks taking into account environmental impacts associated with the installation and operation of water distribution networks. A methodology is established to compute these environmental impacts based on the embodied energy of the different materials used during the construction of water networks. The third model involves a multi-objective optimization approach to the optimal design and operation of water distribution networks considering conflicting objectives. The achieved results prove the capacity of the approach in dealing explicitly with conflicting objectives, with environmental impacts and with future uncertainty. Subsequent to the review and discussion of these proposals, it is presented a set of conclusions and suggestions for future work. viii Resumo Os sistemas de abastecimento de água são infraestruturas cruciais para qualquer sociedade desenvolvida. Nas últimas três décadas, o dimensionamento otimizado destes sistemas tem sido estudado pela comunidade científica e pelas próprias indústrias de abastecimento de água, usando valores de referência para as condições de funcionamento e operação no futuro. As soluções ótimas obtidas em tais condições poderão funcionar inadequadamente se a realidade vier a mostrar-se consideravelmente diferente do previsto. Em condições reais, os sistemas de abastecimento podem ser expostos a situações críticas de operação como a rotura de condutas e reservatórios, a falta de energia ou alterações drásticas nos consumos. Apenas uma atitude proactiva que considere os riscos desde o início do projeto até ao final do ciclo de vida das redes permite lidar com situações de incerteza. As entidades que gerem estas infraestruturas são responsáveis pelo planeamento e operação das redes devendo agir no sentido de mitigar os riscos. Porém os investigadores podem auxiliar a justificar e a tomar boas decisões através do desenvolvimento de ferramentas de otimização. A contribuição central deste trabalho encontra-se nas ferramentas de apoio à decisão desenvolvidas para a otimização do dimensionamento e operação de redes de distribuição de água capazes de ter em conta a incerteza e de encontrar soluções robustas. Trata-se de infraestruturas onerosas em que, uma vez construídas, dificilmente se consegue alterar as suas condições físicas sem incorrer em grandes investimentos. Considerar a incerteza no processo de decisão permite definir soluções fiáveis para um longo horizonte de projeto. Nos capítulos iniciais, este trabalho refere alguns conceitos e definições relativos a redes de distribuição de água. É realizado o estado da arte sobre técnicas de otimização e metodologias de apoio à decisão. São também apresentados dois algoritmos de otimização, ix que posteriormente serão aplicados à resolução dos modelos propostos: um algoritmo baseado no método de recozimento simulado para resolução de prolemas uni-objetivo e um outro algoritmo baseado no método de recozimento simulado para problemas multiobjectivo. De seguida são propostos modelos de apoio à decisão passíveis de serem usados para melhorar as soluções a implementar no âmbito das redes de distribuição. Primeiro é descrito um modelo de otimização robusta de sistemas de distribuição de água sujeitos a diferentes condições de operação. Os resultados indicam que a robustez pode ser incluída de formas diferentes e com diferentes níveis de fiabilidade, através de pequenos incrementos nos custos. Seguidamente é proposto o contributo principal deste trabalho: um grupo de três modelos de apoio à decisão baseados no conceito de opções reais. O primeiro trata da otimização do dimensionamento e operação de redes de distribuição de água considerando a incerteza ao longo da vida de funcionamento da rede. Os resultados mostram que a inclusão de flexibilidade no processo de decisão possibilita definir soluções mais adaptáveis à medida que surge nova informação e possibilita evitar problemas