This paper summarizes practical experiences of solving the UC problem with AC optimal power flow constraints using three main approaches: 1) solving a MINLP with commercial solvers, 2) an outer approximation approach with Successive Linear Programming (SLP), and 3) a Second Order Cone Programming (SOCP) approximation. We show a comprehensive review of each approach, including main characteristics and drawbacks. Although non-linear solvers have improved their performance in recent years, our results suggest that the binary nature of the variables in the UC problem still increases the solution time up to 75 times in comparison to a relaxed version of the UC. SLP approaches have shown a good behavior finding integer feasible solutions, however, the solution times are almost twofold the ones obtained for the MINLP. SOCP is a promising approach because it improves almost 6 times the approximation made by the classic DC approach. Nevertheless, its solving times are comparable to those found using the MINLP. Finally, parallel computing techniques could improve the performance of each approach in order to make them more computationally efficient for large-scale UC problems.
Keywords: Unit commitment, AC optimal power flow, Mixed Integer Non-Linear Programming, Second Order Cone Programming, Outer Approximations
Publicado: junio 2019.