Applications of robust failure detection algorithms to power systems

As modelling and simulation become increasingly popular in the design process and as an alternative to expensive testing, fault detection methods based on model identification algorithms become more reliable as well as less expensive and easier to implement. In this paper we discuss the application of two active fault detection algorithms based on model identification to power systems. The algorithms are similar in theory though differ in implementation. The first is a direct optimization approach that handles more general systems and more varied constraints. It requires more sophisticated software but it's easily adapted to more than two models. The second algorithm is a constrained control approach that can be implemented on common math software, such as Matlab or Scilab, and handles model uncertainty. In both cases, the algorithms are free of false alarms depending upon the quality of the models used.