Techniques for Managing Grid Vulnerability and Assessing Structure

Abstract:

As power grids increasingly rely on renewable power sources, generation fluctuations play a greater role in system operation. Even small variations may cause transmission lines to overheat, provided they push the system operating point in an undesirable direction. A multi-time-step optimization framework is proposed for quantifying system vulnerability to renewable fluctuations. An efficient solution algorithm for the resulting quadratically-constrained quadratic program (QCQP) is also presented. Case studies demonstrate the method’s effectiveness for anticipating line temperature constraint violations due to small renewable generation shifts. The method can also identify efficient generator dispatch adjustments for cooling an overheated line, making it well-suited for power system operation.

The other thrust of this thesis pertains to network and time series data characteristics. In particular, the network graph’s maximal clique composition is studied. This property has performance implications for the semidefinite relaxation of the optimal power flow problem. It is well known that performance improves when the semidefinite constraint is decomposed according to the underlying network’s maximal cliques. Further improvement is possible if some cliques are merged, trading off between the number of decomposed constraints and their sizes. The potential for improvement over existing clique-merge techniques is explored.

Chair: Professor Ian Hiskens