Fault diagnosis and prediction on uninterruptible power systems

The authors deal with a fault diagnosis technique, particularly suitable for power electrical devices, based on the integration of simulation and identification methods. The device under analysis was simulated in faultless and faulty conditions and the experimental validation was carried out. Both the system parametric "unfaulty" model and a set of fault models, in terms of model parameter variations, are identified by means of a recursive algorithm working on the device input/output signals generated by the simulation program. The experimental results, concerning the application of this technique to diagnose failures on a 2 kVA monophase uninterruptible power system are reported