摘要：AbstractA procedure based on trigonometric filters is proposed for parameter identification in a hydroelectric power generating unit approximated by a Multi Input - Single Output linear continuous-time varying system. First the discrete cosine transform is used as an interpolation filter and secondly an integral transform based on a Fourier kernel is applied to obtain a discrete-time model. Then a recursive least squares algorithm is performed to estimate abrupt changes of parameters and thus detect faults in the power plant. The whole method is successfully applied to a scenario fed with real sampled data featured by a poor excitation.
