Cooperative algorithms and abductive causal networks for the automatic generation of intelligent substation alarm processors

To improve the overall power system network maintenance, a challenging subject is to have a day-by-day interpretation of the alarm messages produced in every substation. Our research goal is to design a system able to automatically produce a set of local alarm processors (one for each substation). We address the problem of transforming the substation-dependent knowledge of the ISAP prototype-the substation model and diagnostic mechanism-tightly coupled into the rule base to a structural scheme of knowledge loosely coupled to the diagnostic system. We describe DX, a general diagnostic problem solver based on a qualitative abductive model and the set covering optimization paradigm. The system follows the asynchronous team approach: it is formed by a set of parallel unsupervised cooperative algorithms which work on populations of solutions