Description | François Bouffard Assistant Professor Department of Electrical and Computer Engineering McGill University Recent work on umbrella constraint discovery (UCD) has shown great promise in streamlining the solution of security-constrained optimal power flow (SCOPF) problems. The solution of the UCD problem itself is not trivial, however. In this talk, we present a significant, yet simple, improvement to the decomposition approach used to solve UCD. This improvement exploits the inherent structure of the parent SCOPF problem. Moreover, given the promising results from UCD-SCOPF, we have moved on to apply UCD to the solution of the classic thermal unit commitment problem with the hope that it could identify strong branching and time decoupling opportunities. We prove that, unlike SCOPF, the unit commitment problem does not have a favorable structure for which UCD can be effective. From these experiences, we can draw conclusions on the desirable features of mathematical programs for which UCD can be most effective, and we make recommendations for other important electricity generation planning problems. |
---|