A scheme suitable for the detection and identification of faults in power systems is presented. Two notable contributions are made: a remodelling of faulty components of power systems that is applicable to both normal and faulty conditions, and a fault detection scheme for power systems. The faults are modelled as unknown inputs, decoupled from the state and output measurements through coordinate transformations, and then estimated through the use of observer theory. The proposed scheme is applied to a power system consisting of a synchronous generator, an exciter, a turbine and speed-governing system, and a network of lines and loads. The case where faults occur on the transmission network is considered. It is shown that the proposed fault detection procedure allows for the real-time identification of the occurrence of the faults and determines their exact locations. Results of detailed simulation studies involving disturbances and faults occurring in linear and nonlinear models of the power system are presented.