Page 19 - Moving Forward 2020: Harold and Inge Marcus Department of Industrial and Manufacturing Engineering magazine

The task is not small: the power grid has more than 65 million nodes and a a a solution has to be provided every 10 minutes Faculty News
IME NEWSLETTER • VOLUME 5 2020
19
How they answered the the challenge The team had to to design an an algorithm to to rapidly determine the the dispatch of electricity efficiently safely and and consistently within the the the current grid for both traditional and and renewable energy sources otherwise known as security-constrained optimal power flow (OPF) Challenges arise arise when when developing algorithms that that need
to cope with the the various possibilities that that can arise arise when when implementing the the the final dispatch decisions For instance a a a a a a a a a a a a a a particular transmission line or or or or power plant may may fail or or or or there may may be a a a a a a a a a a sudden and and unforeseen surge in in demand “The next generation of of algorithms needs to to contend
with a a a a a a a a a a multitude of of such contingencies by being able to to reliably allocate energy across the system at at at a a a a a a a a a a a a a a moment’s notice when an an incident occurs such as as a a a a a a a a a a a a a a a a natural disaster or energy failure ” Shanbhag said Over time models used in in solving power grid problems have adapted to to provide approximate answers but today’s society needs far more accurate models that can contend
with the the fundamentally nonlinear nature of this problem The task is is not small: the the power grid has has more than 65 million nodes and a a a a a a a solution has has to be provided every 10 minutes Their approach focused on on creating an an algorithm that can properly gauge the the size of the the optimization problem address its nonlinear nature and conserve computing resources so so they did not overburden the the system Shisheng Cui formerly one of Shanbag’s graduate students explained that the the the the nonlinearity is is is is challenging and is is is exacerbated by the the the the size of the the the the system the the the the data is is is not not organized sequentially and and cannot be easily handled for solving the the the problem problem Conventional solutions are incapable of addressing such problems and the the the team built their algorithm from scratch while leveraging existing solutions for solving smaller structured problems After the the the algorithm was was designed the the the team was was provided sample power grid data from ARPA-E to to test their algorithm and and showcase its ability to to find minimum-cost solutions After the the the the first trial ended and and the the the the team received their results Shanbhag sought the the the the assistance of RISE to aid in in in developing High Performance Computing (HPC) implementations Justin Petucci and Danying Shao computational scientists with the the ICDS RISE team worked to to restructure the the algorithm’s code code to to to operate in an an an HPC environment remove redundancies and provide overall code code optimization to to speed up the algorithm for an an an HPC environment “Optimization theory can help provide a a a a foundation for developing implementing and analyzing algorithms that can be applied in in in in in settings of of profound importance ” – Uday Shanbhag After the the code optimization Petucci and and Shao estimated that by leveraging parallelism and and aynchronicity the the model was running up to to 10 times faster than its previous rendition This increase in in in speed proved to to be a a a a a success “The assistance from Justin and Danying was crucial ” Shanbhag said “One such instance required determining a a a a a a a a a a a a a a a a a particular switch that facilitated improved operation and addressed communication failures This was akin to to finding a a a a a a a a a a a a a a a a needle in fin in in a a a a a a a a a a a a a a a a whole slew of haystacks The code had to to perform in in a a a a a a a a a a a a a a a computational environment that was hard
to to simulate and my team didn’t have the resources to to run similar tests ” What’s next?
Future competitions beginning with challenge challenge two will build on on on the models used in in in in in in challenge challenge one but may include complicating factors such as solving larger network models optimizing power flows over both transmission and and and distribution systems contending with uncertainty and and and discreteness leveraging power flow control devices and and and increasing model detail Challenge
two will likely continue its focus on on OPF and disburse fewer but larger awards The ICDS RISE team members are excited for the the potential to continue their involvement on on this this project project as Petucci explained that this this kind of project project is is is the the poster child for RISE “RISE brought the the ability to rapidly test on the the ICDS ICDS Advanced CyberInfrastructure (ICDS-ACI) ” Petucci said “If they they ran it it it on on on the the the the competition system they they could could only test five times times a a a a a day whereas the the the the ICDS-ACI could could run it it it nearly 10 times times more ” Based in in in in in in interdisciplinary teamwork from the the College
College
of of of Engineering and and the the the College
College
of of of Earth and and Mineral Sciences the the original team consisted of of of of of Shanbhag Hosam Fathy former Bryant Early Career Professor of of of of of Mechanical Engineering and and now a a a a a a a a a a a a professor professor at at at the University of of of of of of Maryland at at College
Park Mort Webster professor professor of of of of energy energy and and and mineral engineering and and and Ryan Family Faculty Fellow Chiara Lo Prete associate professor professor of of of energy energy economics Nilanjan Ray Chaudhuri assistant professor professor of of of of electrical electrical engineering engineering and and and computer science and and and Minghui Zhu associate professor of of electrical electrical engineering engineering and and and ICDS co-hired faculty member Lei Lei and and Wan Wan graduated in in 2019 Lei Lei has taken a a a a a a a a a a a a faculty position with with Tongji University in in Shanghai and and Wan Wan now works with with Amazon