The objective of this research is to develop improved tools for planning for electric power systems. The approach of this research is to combine analysis at hourly and annual timescales so that constraints on how electricity generators operate are accounted for in long-term investment planning for generation technologies. Such constraints are particularly important with increasing use of renewables, storage, and responsive demand. The project will apply approximate dynamic programming and traditional integer optimization techniques to explore decisions under uncertainty in both time scales. The structure of the operations and investment sub-problems will be exploited to develop efficient methods for optimizing the full system, accounting for uncertainty in demand, renewable generation, fuel prices, and possible environmental regulations.
Intellectual Merit This project will develop new methods for optimizing large engineering systems under uncertainty by developing new algorithms and data structure designs. It will also advance the state-of-the art for multi-timescale decision models where the smaller timescale is computationally expensive.
Broader Impacts This work will significantly improve planning for advanced electric power systems that combine renewables and other advanced technologies to meet environmental and energy requirements, and will identify system designs with lower costs and greater resiliency. The methods developed will be usable by power companies, independent system operators, and variety of governmental and non-governmental agencies that are in the process of designing the next-generation power system. This project will also provide education and training to undergraduate and graduate students, including women and underrepresented minorities, in operations research and power systems modeling.
" (PIs Webster and Marzouk), funded by the National Science Foundation Award number 1128147. In this project, we have developed a large set of methods and computer software all focused on how to design and operate the electric power grid of the next century. Given the enormous uncertainty in future environmental regulation (including carbon emissions), future energy technology costs and performance, and future fuel prices, planning the needed investments in the power system will necessarily be a decision made under uncertainty, with future decisions adapting to new information about these uncertainties as we learn. In this context, near-term decisions should be robust and flexible. However, the computational tools needed to perform this kind of planning do not currently exist. At present, many industrial and government participants in the system perform simpler analyses that ignore uncertainty, or run scenarios that do not help to identify near-term robust decisions. Moreover, it is likely that the future electric grid will have much more of its generation coming from wind and solar power, which can vary rapidly over the course of the day. Planning the needed investments in other types of generation requires more detailed modeling of the hour-to-hour behavior over years, which was previously not possible. This project has developed a number of novel techniques that will enable the generation of computer tools to be used by industry to explicitly plan for uncertainty in choosing what new power plants and what transmission lines should be built. The computer tools developed and now available to others from this project will also allow for planning future power plants that are required to offset the variability of energy from wind and solar to maintain the reliability of the grid.