Access to safe, clean drinking water is a major challenge for many communities. These communities are often near seawater and/or brackish groundwater sources, making desalination a possible solution. Unfortunately, desalination is energy intensive and a reliable, inexpensive power supply is also challenging for remote locations. Photovoltaic reverse osmosis systems (PVRO) can be used to provide water for underserved communities. A feasibility study which demonstrates the economic viability of such systems is discussed here.

PVRO systems are assembled from mass-produced modular components. This approach reduces manufacturing costs. However, designing a system optimized for a specific location is difficult. For even a small inventory of components, the number of design choices is enormous. A designer with significant expertise is required to tailor a PVRO system for a given location, putting this technology out of reach of many communities.

This thesis develops a modular design architecture which can be implemented in a computer program to enable non-experts to configure systems from inventories of modular components. This architecture is not limited to PVRO systems, but can also be used to design other systems composed of modular components such as cars, electronics, and computers. The method uses a hierarchy of filters to limit the design space based on design principles and calculations. The system is then configured from the reduced design space using optimization methods and detailed system models.

In this thesis, the modular design architecture is implemented for PVRO systems. A set of detailed physics-based system models are developed to enable this process. A novel method of representing a PVRO system using a graph is developed to enable rapid evaluation of different system configurations. This modeling technique is validated using the MIT Experimental PVRO system constructed as part of this research.

A series of case studies are conducted to validate the modular design approach for PVRO systems. The first set of case studies considers a deterministic solar input and water demand. The design goal is to determine the lowest cost system that meets the water demand requirements. It is shown that the method is able to tailor systems for a wide range of locations and water demands from a large system inventory. The validity of these solutions is demonstrated by simulating a custom designed system in the wrong location. Another case study shows that the approach can be used to determine market potential of new components.

The second set of case studies considers variations in the solar radiation and water demand. The design goal is to determine the lowest cost PVRO system that meets the water demand profile with a specified probability. Two methods that use historical solar insolation and water demand to account for variations are presented. The first method characterizes the historical data and develops models to synthetically generate solar insolation and water demand profiles, and then simulates the system performance over 100 years to calculate the loss-of-water probability. In the second method, distributions of solar radiation and water demand are calculated from historical data and used to directly calculate the probability of running out of water in the worst month of the year. Both methods are implemented and shown to produce feasible system configurations. The direct calculation method is shown to reduce the required computation time and is suitable for different systems with variable inputs.