Shifting energy consumption into times when renewable energy is abundant can increase renewable integration and reduce greenhouse gas emissions. As large energy users with flexibility in their operations, water and wastewater utilities can participate in this type of energy demand management. However, the primary goal of the water sector is to provide clean, safe, and adequate drinking water and protect human health and the environment by treating wastewater to appropriate standards. Water or wastewater treatment and delivery systems are both complex, with their own operational challenges; optimizing energy use must not interfere with the system's primary goals. In this dissertation, I present research focused on exploring techniques that allow water and wastewater utilities to perform energy demand management while still prioritizing key operation performance metrics such as water quality or system reliability. I also leverage data analytics, machine learning, and visualization to allow users, including water and wastewater utility optimizers, to better characterize and improve optimization problem formulations and methods. The first investigation focuses on testing load shifting strategies at a full-scale wastewater treatment plant to participate as a demand resource. During these test events, the facility shifted energy load by modifying select operations with little to no impact on water quality. A cost-benefit analysis showed that the facility achieved cost savings of up to 4.8% by participating in the proxy demand response program, which allows users to bid on the California Wholesale energy market. From this case study, we identified two main barriers to wastewater utilities participating as a demand resource: the difficulty of correctly timing energy reductions to demand response periods and the inability of the standard baseline methodology to measure demand reduction for this complex system. In the second research effort, we present a new optimization problem formulation for the water distribution system pump optimization problem: secondary time-based controls. This proposed pump control uses a hierarchal structure to allow utilities to prioritize maintaining sufficient water reserves while responding to time-based energy incentives. The formulation was tested on three case studies and compared to two baseline pump control decision variable representations. A Monte Carlo sensitivity analysis was also conducted to determine the robustness of the control structures to uncertain water demands. Results show that this formulation similarly or further reduces energy costs than the two benchmark decision variable formulations without reducing average storage or violating operational constraints. In addition, the secondary time-based controls more consistently maintain water reserves and prevent constraint violations with uncertain demands, allowing water utilities to more comfortably manage energy demand to support renewable energy growth. In the final study, we developed a visual analytic framework that characterizes the optimization fitness landscape to help users improve the optimization problem formulation and search efficiency. We also present a corresponding optimization method that guides the optimization using similar interpretable machine learning methods. Both the framework and optimization method were designed for use in the water distribution pump operation problem, but many of the analytic and visualization techniques could be applied to a broad range of complex optimization applications. We used the framework to examine the differences in fitness landscape between two different decision variable problem formulations on a benchmark water distribution system. We then tested the performance of several existing heuristic optimization methods either with or without guidance from the visual analytic framework and compared them to the proposed optimization method. The existing methods informed by the visual analytic framework and the new optimization method showed improvement over standard optimization.