TY - GEN
T1 - Network flow and copper plate relaxations for AC transmission systems
AU - Coffrin, Carleton
AU - Hijazi, Hassan
AU - Van Hentenryck, Pascal
N1 - Publisher Copyright:
© 2016 Power Systems Computation Conference.
PY - 2016/8/10
Y1 - 2016/8/10
N2 - Convex relaxations of the power flow equations and, in particular, the Semi-Definite Programming (SDP), Convex Quadratic (QC), and Second-Order Cone (SOC) relaxations, have attracted significant interest in recent years. Thus far, little attention has been given to simpler linear relaxations of the power flow equations, which may bring significant performance gains at the cost of bounding strength. To fill the gap, this paper develops two intuitive linear relaxations of the power flow equations, one based on classic network flow models (NF) and another inspired by copper plate approximations (CP). Theoretical results show that the proposed NF model is a relaxation of the established nonlinear SOC model and the CP model is a relaxation of the NF model. An experimental study on 94 real-world test cases demonstrates that the linear relaxations are very fast and only reduce the bounding strength by a few percent. Consequently, considering the linear NF and CP relaxations alongside established relaxations (SDP, QC, SOC) provides a rich variety of tradeoffs between relaxation strength and performance.
AB - Convex relaxations of the power flow equations and, in particular, the Semi-Definite Programming (SDP), Convex Quadratic (QC), and Second-Order Cone (SOC) relaxations, have attracted significant interest in recent years. Thus far, little attention has been given to simpler linear relaxations of the power flow equations, which may bring significant performance gains at the cost of bounding strength. To fill the gap, this paper develops two intuitive linear relaxations of the power flow equations, one based on classic network flow models (NF) and another inspired by copper plate approximations (CP). Theoretical results show that the proposed NF model is a relaxation of the established nonlinear SOC model and the CP model is a relaxation of the NF model. An experimental study on 94 real-world test cases demonstrates that the linear relaxations are very fast and only reduce the bounding strength by a few percent. Consequently, considering the linear NF and CP relaxations alongside established relaxations (SDP, QC, SOC) provides a rich variety of tradeoffs between relaxation strength and performance.
KW - Convex Optimization
KW - Linear Programming
KW - Optimal Power Flow
KW - Power Flow Relaxation
UR - http://www.scopus.com/inward/record.url?scp=84986552416&partnerID=8YFLogxK
U2 - 10.1109/PSCC.2016.7540869
DO - 10.1109/PSCC.2016.7540869
M3 - Conference contribution
T3 - 19th Power Systems Computation Conference, PSCC 2016
BT - 19th Power Systems Computation Conference, PSCC 2016
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 19th Power Systems Computation Conference, PSCC 2016
Y2 - 20 June 2016 through 24 June 2016
ER -