A Kalman filter based method is proposed for the harmonics analysis of three-phase grid signals. The method sets up two Kalman filters

one of which works for the estimation of harmonics' amplitude and phase

and the other for the estimation of fundamental frequency. The relationships between these two Kalman filters are described. The state estimate of the former can be used to calculate the measurement of the latter and the state estimate of the latter provides model frequency to the former. To accelerate the algorithm

the initialization of Kalman filters and an algorithm for detecting the abrupt change in signals are presented. It is verified by simulation that the proposed method is still highly efficient with abrupt changes in three-phase grid signals

in amplitude

phase and frequency

with continuous changes in amplitude and frequency

and under unbalanced conditions. The experiment applies the proposed method to the voltage and current signals of a three-phase motor under non-ideal grid conditions

where the mean relative errors between the reconstructed signal and the original one are only 0. 67% and 1. 04%

respectively. This work is supported by Zhejiang Provincial Key Research and Development Program of China(No. 2017C01039)and National Natural Science Foundation of China(No. 51377142).