Overcoming modeling barriers in long-term interdependent infrastructure systems planning

Long-term planning in water infrastructure involves increasingly competing financial, regulatory, and environmental constraints and objectives. To face the challenges of aging infrastructure, emerging contaminants, climate change and natural hazards, opportunities for long-term, multi-objective water system optimization modeling have never been greater. The intent of this thesis is to confront the technical challenges of implementing such models in standard planning practice, and to demonstrate their potential to generate insights in a real case study system.A groundwater utility is supplied by several hundred spatially distributed, small-scale pumping stations. Chapter 2 studies the impact of mathematical model formulation on performance and decision flexibility in the least-cost pumping and treatment scheme. Several high-level linear and mixed-integer linear program (MILP) formulations are presented, illustrating the analytical challenges inherent in this type of problem, as well as the value of different levels of decision flexibility and model solvability. The heuristic and MILP techniques applied, as well as cost regression functions developed utilizing the U.S. EPA’s Work Breakdown Structure (WBS) models, produce insights about economies of scale in the technologies considered and degree of treatment centralization, and the expected range of objective performance that could justify use of these techniques in planning practice. However, the problem simplification required to formulate the MILPs and lack of optimal substructure to improve solvability, along with the similar cost performance of different solution alternatives, limit the effectiveness of exact techniques to guarantee a meaningful comparison of planning alternatives.Chapter 3 develops two nonlinear models for the case study to demonstrate the importance of treatment facility design and operational parameters that that were inflexible in Chapter 2’s models, in terms of three planning objectives: costs, service continuity, and longterm water quality management. A flexible modeling framework is developed with customized versions of the WBS models and implemented within an existing evolutionary algorithm. The chapter discusses barriers and benefits to using a complex modeling framework as a long-term planning tool: namely, the resources needed to build and maintain it, versus its reusability, allowing development of external modeling tools to improve its usefulness over time.One fast-growing application of such modeling capabilities is system-of-systems (SoS) modeling for community resilience planning. Recent growth in resilience modeling creates both an opportunity and a challenge for systems modelers to apply operations research (OR) ideas when integrating interdisciplinary infrastructure system models while avoiding ad hoc, methodology-driven approaches. Chapter 4 proposes a planning-driven approach to systematically define resilience metrics that align the model building process with system and community resilience goals. Chapter 5 applies the approach to evaluate system resilience metrics in the case study system. A modeling framework is developed to simulate the damage and recovery of physical and human components in a drinking water system and its supporting electric power and transportation networks under two natural hazard scenarios. The chapter compares the effectiveness of the selected metrics and implications of their use as objectives in a resilience planning context.Model formulation, problem representation, model capabilities, and objective definition are only a part of the challenge to using optimization models for long-term water planning in practice. Many public utilities use decision staging to separate decisions or objectives, either to intentionally limit the planning scope or because the stakeholders and timeframes for certain decisions are disconnected.