Research on Harmonic Detection Based on Improved Kalman Filter and Wavelet Packet Transform

Abstract

Harmonic signal will be mixed with some noise and singular signals in the detection process. Aiming at the disadvantage that Kalman filter is only applicable to Gaussian noise, this paper proposes an algorithm combining wavelet packet transform and improved Kalman filter algorithm. Firstly, the outliers in the original signal are detected and removed by wavelet transform to prevent the offset or even divergence of the Kalman filter. Kalman filter optimally estimates the processed noisy signal. In this process, according to the residual outliers in the signal, a threshold is designed to adaptively correct the Kalman gain to ensure that the noise is filtered out while retaining high-order harmonics. In the process of detection, the noise is prevented from submerging the high-order harmonic. Finally, the high resolution of wavelet packet transform is used to realize the accurate detection of harmonics. Compared with traditional threshold filtering, it has obvious advantages. The results show that the method proposed in this paper can realize accurate harmonic detection. This is of great significance for determining harmonic source, harmonic control and maintaining power grid security.