An abductive inference model for alarm processing [power system fault diagnosis]

This paper introduces a system modeling concept and a new algorithm based on abductive inference for handling diagnostic problems in electrical power systems. The overall technique relies on the inherent functional and logical relationships between system elements such as lines, buses, relays and breakers to generate a diagnostic model rather than on experientially derived heuristics. This approach allows for a systematic implementation scheme that is more readily adapted to any power system. Logical models for relays, breakers, lines and buses are presented along with examples on how these models can be interconnected to represent a power system. The resulting system model can simulate the "forward" behaviour of a power system. Inputs to the forward model consist of disorders or disturbances, and the outputs are manifestations or alarms. Using abductive logic, a "reverse" system model can be generated. This system model takes the alarms as inputs and produces a diagnosis by generating sets of disorders or disturbances that explain the observed alarms.