From Static to Dynamic Observability for a Multiple-Model Filter Based Fault Diagnosis Approach

In a large power system computational complexity is a challenge for applying model-based fault diagnosis techniques, such as multiple-model filtering (MMF) approach. A common strategy for containing complexity in a system is to partition the system model into subsystem models. The authors have proposed a partition algorithm for large power systems [1] to tackle the computational complexity problem for fault diagnosis using MMF approach. A partition is defined as a partial circuit of a power system surrounded by a group of PMU-like sensors that provides full static observability of the electrical network. In the multiple-model approach a time domain dynamic model is built for each possible topology in a partition with nominal expected load value. As these models are in time domain, dynamic observanility is a requirement for detectability of each partition's mode. In this paper authors proves that their proposed partition algorithm upholds the dynamic observability of all the partial circuits under the assumptions of full system static observability, a lumped circuit approximation of transmission lines, and a circuit parameter ratio condition.