Robust Topology Detection Under Load Uncertainty

A multiple-model based detection method for power system topology is developed to operate under substantial load uncertainty. Using a lumped transmission circuit model for each anticipated topology and assuming synchrophasor availability at all generator terminals, a convex hull of potential load current perturbations is mapped into a decision space spanned by vectors of measurable current synchrophasors, where set convexity is preserved. A robust detector design algorithm is presented to find the set of detection hyperplanes that fully separate the topology-defined sets of measurable current synchrophasors mapped from the largest convex hull of load perturbations. An optimally robust detector is designed for the IEEE 9-bus test system, demonstrating a much larger tolerable load uncertainty compared with a recently reported norm-based topology detector. Our detector displays robustness under a uniform load uncertainty of ±0.21 per unit across the real and reactive current phasor components, or an uncertainty equal to ±29% of the nominal load. The method features several scaling factors, allowing the designer to tolerate higher uncertainty on a specified load at the cost of less tolerance elsewhere.