Fault diagnosis based on partitioned power system models

Through partition of the model of a full power transmission network, this paper resolves the complexity issue arising from using a multiple model filtering (MMF) approach to transmission short-circuit fault diagnosis. The MMF approach to power system diagnosis was recently developed to mitigate cascading failures in power systems, caused by protection misoperations, through providing decision support for protective control actions. The partition is defined by the structure of a sensor network overlaid on the transmission network to support, among many functions, fault diagnosis. The ability to diagnose faults using the MMF approach relies on the observability of independent current and voltage variables of the transmission network from time-synchronized measurement samples acquired by the sensor network. The paper shows that such observability is retained in all partitioned networks, for which the complexity of real-time computation and communication is bounded despite the size of the full network. A set of rules for partitioning is stated in graph theoretic terms. The MMF approach to diagnosis of transmission line faults is applied to the IEEE 68-bus test system and the benefit in complexity reduction through partition is quantified.