Transmission fault diagnosis with sensor-localized filter models for complexity reduction

This paper presents the results of development, implementation, and test of a sensor-defined partition algorithm of a transmission network, and its application to real-time diagnosis of transmission circuit faults using a multiple-model filter (MMF) method. The partitions are bordered by the measurement nodes of an advanced sensor network overlaid on the power transmission network. The partition aims to reduce the complexity growth of computation and communications from exponential to linear with respect to the system size, while fulfilling high-speed and high-precision diagnosis. The diagnosis is a part of a secondary protection designed to cost-effectively mitigate primary protection misoperations and to deal with new power flow patterns of the evolving power grids. The paper delineates the sensor network requirements that guarantee the transmission networks dynamic observability necessary for high-precision diagnosis with MMFs. A full set of diagnosis filters are built for the 68-bus test system based on the partitioned dynamic models of the transmission network under normal, short-circuited, tripped, and falsely tripped modes, where the partition is defined by a PMU-like sensor network. Diagnosis performance is evaluated through hybrid-simulations, and the reduction of complexity as a result of partition is quantified.