ESPRIT assisted artificial neural network for harmonics detection of time-varying signals

This paper presents a new approach for harmonics estimation of time-varying power supply signals using adaptively trained artificial neural network (ANN). The proposed method employs the high resolution estimation of signal parameters via rotational invariance technique (ESPRIT) that assists ANN to continuously update its parameters according to the varying input signal to provide more accurate and reliable estimates of harmonic amplitudes. New ESPRIT assisted online training scheme makes the neural network based harmonics estimation techniques more versatile for stationary as well as time-varying power supply signals. The performance of the proposed method is validated on the time-varying synthetic signals with radial basis function neural network.