On Identifiability Of Sparse Gross Errors In Power System Measurements

In this paper, we present feasibility study of sparse error correction in power system measurements. Legacy bad data detection mechanisms have been shown to be prone to make erroneous decisions when elaborately designed errors are injected by cyber attacks. In order to effectively handle such gross errors, sparse error correction framework has been suggested in the literature. For proper utilization of sparse error correction framework, it is important to understand what are the gross errors that are fundamentally identifiable based on the sparsity assumption. Such understanding can help us find optimal security resource allocation that can restrict the gross error locations such that the errors are identifiable. In this paper, we present a few conditions that can guarantee that gross errors are identifiable. In addition, we provide sufficient conditions under which a summation of identifiable gross errors with disjoint supports is also identifiable. These condition can be used to verify identifiability of gross errors in the measurement system by checking local conditions of smaller subsystems.