A, neural learning approach for tlime-varying frequency estimation of distorted harmonic signals in power systems

In this paper, we consider the problem of estimating the frequency of a sinusoidal signal whose amplitude and frequency could be either constant and time-varying. We present an artificial neural network approach for the on-line estimation of the signal frequency. The neural network architecture and learning is formulated based on an original decomposition of the signal to estimate. We show that the neural estimator can be implemented using hardware technologies and can be efficiently be compared to conventional frequency estimation algorithms. The problem of detecting frequency variations in a power system is addressed and the results show that the neural frequency estimator is efficient. Simulation and experimental examples on a real-time platform are included to show the performance in terms of both estimation and detection