Researchers from the University have received a $1.4 million grant from the Department of Energy to help develop data on power system optimization in energy grids.
The team will work to develop new test cases to formulate better software algorithms for transmission operators to run the energy grid — algorithms which regulate energy amounts. These operators are largely non-profit government agencies. The need for such research stems from the ongoing energy transition from traditional, emission-heavy sources such as coal and nuclear power to cleaner, renewable sources like wind and solar.
Ian Hiskens, professor of electrical engineering and computer science, said the energy transition has several impacts on the electric grid.
“The main issue with renewables is variability,” Hiskens said. “Output for a coal power station can be fixed and controlled. If you deal with a wind farm, you have no control over the output. The energy grid needs to be upgraded to manage fluctuations and variability.”
Pascal Van Hentenryck, professor of industrial and operations engineering and the leader of the project, said the change alters fundamentals about how energy grids function.
“As we push the frontier as to how much renewable energy can be put in the network, the basic assumptions of the grids are no longer true,” Van Henternyck said. “We need to change the way the system operates by altering the algorithm that determines the balance of electricity in different places.”
As an example of an alteration, he cited a possible need to import and ship solar energy from sunny places like Arizona to elsewhere in the United States.
Most existing test cases that transmission operators draw from have become “toy problems”, Van Hentenryck said, making them not necessarily reflective of today’s complex networks.
To mitigate this dearth of realistic data sets, the team will embark on a multi-phase two-year project, one of the seven funded by the $11 million Advanced Research Projects Agency-Energy (ARPA-E) program run by the Department of Energy.
The first phase of the project involves data acquisition and modeling. This includes monitoring the life of existing power networks to obtain a year-long series of data. Since retrieving real data from the United States is difficult due to security concerns, the group plans to partner with French utility firm, Réseau de Transport d’Électrique. The data from French and European networks then have to be modeled into an intelligible format.
For the modeling phase, Hiskens, a former power engineer in Australia, said he hopes to use his knowledge about realistic power system domains.
“We are interested in building data sets with numbers, parameters and variables that are sensible when evaluated from the perspective of a real power system,” Hiskens said. “I will be involved in making sure that formats and structures developed are sufficiently rich to capture the idiosyncrasies of power systems.”
However, he noted the difficulty in capturing the uniqueness and intricacy of power systems.
“Ultimately we need to establish a format that is complex, flexible yet useable. This way, others who want to work on power system optimization can test new algorithms based on the data sets, building a community pool of information.
Given the sensitivities of using real data, Van Hentenryck described the challenge of hiding some data and interpreting the overall picture will also be a challenge.
“We will have to use the algorithm to obfuscate data that is sensitive, such that the real data is unrecoverable,” Van Hentenryck said. “Sometimes we get aggregated consumption profile, meaning that we need to desegregate the data into as fine-grained as possible.”
The next phase of the project would be to generate synthetic benchmark data that are not identical, but similar to reality. Subsequently, the team will have to validate its processes and findings, as well as place the test cases on an accessible portal.
The University team will also work closely with researchers from the Los Alamos National Laboratory, which has a long history of providing capability and support to the federal government in modeling and simulating power systems.
Russell Bent, who works as a scientist in Los Alamos, said industry-academia-government collaboration is highly important in improving the United States' power systems.
“My role at Los Alamos is to develop underlying schemas that describe all the components of power systems in their complexity,” he said. “We hope to take a closer look at some of the synthetic problems and test cases being created and make sure there is a realistic flavor to them.”
The ultimate goal, Bent said, is to build better, more efficient algorithms and computational methods for federal customers — the transmission operators.
Van Hentenryck said the grant will largely go into recruiting and sponsoring manpower for the project. Currently, researchers from Columbia University and California Institute of Technology are working alongside the University team. He said he hopes to assemble a team of 15 to 20, comprising undergraduate, graduate, Ph.D. and post-doc students.
Though there are multiple milestones that the team is expected to reach, the project’s ultimate success in power optimization is indeterminate at this point, researchers said.
Van Hentenryck likened the project to an exploration phase.
“We know where we are going, but we don’t know what we are going to find,” he said. “So here we are going to generate these test cases, but we don’t really know how well we can solve them or if the solution can be scalable for future networks that incorporate more renewable energy.”
Nonetheless, he said, he is excited about the potential of the project.
“It’s not every day that you can change the world.”