Data-driven technique for robust fault detection in generators

Protection of a synchronous generator presents a very challenging problem because of its simultaneous system connections on three different sides; the prime mover, grid and the source of DC excitation. Generator Model is a very extensive and complex model and model-based fault detection techniques are difficult to implement. For this data-driven techniques can be applied which need only the process data to establish FDD systems. This paper presents application of subspace aided system identification method and robust residual evaluation using the process data directly, to detect actuator faults occuring in synchronous generators.