Multiobjective Evolutionary Approach to Rehabilitation of Urban Drainage Systems Wilmer Barreto 1 ; Zoran Vojinovic 2 ; Roland Price 3 ; and Dimitri Solomatine 4 Abstract: Urban flooding has become a very important and growing issue around the world. In order to maintain an acceptable performance of urban drainage systems, early rehabilitation plans must be developed and implemented. The allocation of funds to support rehabilitation works should be in a certain sense “optimal” in providing value for money. However, this is a highly demanding and not easily achievable task due to the multidimensional nature of the rehabilitation process, especially taking into account conflicting interests. In this respect, multiobjective optimization using hydrodynamic urban drainage models appears to be promising and more reliable than the traditional engineering approaches. Such optimization is used in this paper to evaluate urban drainage rehabilitation scenarios contrasting investment against flood damages. A small-scale rehabilitation problem is posed and solved. The approach has demonstrated the potential use and combination of multiobjective optimization and hydrodynamic models to analyze urban drainage rehabilitation, providing valu- able information for decision makers. DOI: 10.1061/ASCEWR.1943-5452.0000070 CE Database subject headings: Urban areas; Drainage; Optimization; Rehabilitation. Author keywords: Multiobjective; Urban drainage; Optimization; Rehabilitation; NSGA-II. Introduction Despite the considerable knowledge generated in the area of urban drainage planning and management, there are still many challenges concerning the design and maintenance of urban drain- age systems UDS. The different phases of flow in a UDS are traditionally managed separately, including flow in sewers, over- land flood flows, treatment works, groundwater, and receiving waters. Understanding the relevant physical and chemical pro- cesses and using the appropriate model for each phase is of para- mount importance. Model integration with multiobjective optimization techniques opens up a range of options for a holistic approach to UDS management. In earlier papers Vojinovic et al. 2006; Solomatine and Vojinovic 2008, we have present a meth- odology for optimizing the rehabilitation of urban drainage assets including full utilization of the existing system’s capacity prior to undertaking improvement works, reduction of pollution affecting local receiving waters as well as main watercourses and ground- water, reduction of infiltration/inflow to the network and its ad- verse impact on wastewater treatment, minimization of damages due to surcharge affecting properties and the environment, limit- ing the risk of structural failure and damage to other subsurface infrastructure, prevention of pollution of water supply systems due to infiltration of wastewater, and minimization of expendi- ture. Furthermore, the rehabilitation process should allow for an input by stakeholders. However, when decisions have to be taken, the interests of a wide variety of stakeholders make the interpre- tation of drainage information very difficult. The involvement of many interests usually leads to conflicting objectives. In such a case, managers and system planners need effective methodologies and tools that provide a holistic view in order to come up with an optimal trade-off between different interests. Inevitably, this re- quires the use of multiobjective approaches and tools to deal with the conflicts that arise. There is, however, very little work done in applying such ap- proaches for optimizing UDS rehabilitation strategies. In this paper a multiobjective evolutionary approach for the rehabilita- tion of UDS is introduced and demonstrated on a hypothetical UDS. The work presented here also builds on previous research and the results published by the present writers. It shows how to balance the investment costs in rehabilitating the drainage infra- structure with the damages caused by flooding for a multicriteria scenario. The paper explores and compares the performance of two well-known multiobjective genetic algorithms GAs. The comparison of the algorithm is done using cardinality, computa- tional time, hypervolume and epsilon coefficient with respect to the population size and the number of generations. Previous Work Sewerage rehabilitation is a complex task. This process can be structured into four phases. Phase one is aimed at collecting and 1 Ph.D. Research Fellow, UNESCO-IHE Institute for Water Education, P.O. Box 3015, 2601 DA, Delft, The Netherlands; and Dept. of Hydraulic Univ. Lisandro Alvarado, Venezuela corresponding author. E-mail: wbarreto@ucla.edu.ve 2 Associate Professor, UNESCO-IHE Institute for Water Education, P.O. Box 3015, 2601 DA, Delft, The Netherlands. E-mail: z.vojinovic@unesco-ihe.org 3 Professor, UNESCO-IHE Institute for Water Education, P.O. Box 3015, 2601 DA, Delft, The Netherlands; and Water Resources Section, Delft Univ. of Technology. E-mail: r.price@unesco-ihe.org 4 Professor, UNESCO-IHE Institute for Water Education, Delft, The Netherlands; and Water Resources Section, Delft Univ. of Technology. E-mail: d.solomatine@unesco-ihe.org Note. This manuscript was submitted on November 17, 2009; ap- proved on December 14, 2010; published online on December 16, 2009. Discussion period open until February 1, 2011; separate discussions must be submitted for individual papers. This paper is part of the Journal of Water Resources Planning and Management, Vol. 136, No. 5, September 1, 2010. ©ASCE, ISSN 0733-9496/2010/5-547–554/$25.00. JOURNAL OF WATER RESOURCES PLANNING AND MANAGEMENT © ASCE / SEPTEMBER/OCTOBER 2010 / 547 Downloaded 23 Mar 2011 to 194.171.38.2. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org