Solving ACOPF Problems



The ACOPF problem concerns the optimal selection of an operating point for an electrical power grid.  Even though the problem was initially formulated in the early 1960s, it is now gaining increased prominence due to impending changes in the electrical power delivery, which are placing increased pressure on simple approximations now used in operations.  We will begin the talk by describing our participation in an ongoing and previous instances of an ARPA-E-run competition on ACOPF, which uses modern and highly extended (and realistic) formulations for the problem, together with very large-size and realistic data sets with demanding computational requirements.  We will then explore challenges in implementing robust and fast ACOPF solvers using standardized software (the meaning for this statement will be made clear in the talk).  Finally we will conclude with some observations on purely linear formulations for ACOPF.  This is joint work with several authors.



Daniel Bienstock is Liu Family Professor of Operations Research, with joint appointments in Applied Math and Electrical Engineering, at Columbia University.  His work focuses on high-performance algorithms for nonconvex optimization, with focus on large-scale cases; and with an additional focus on problems arising in engineering.  He became an Informs Fellow in 2013 and received the Khachiyan Prize in Optimization in 2022.  He received a PhD from MIT in Operations Research.