Signal Processing and Soft Computing Approaches to Power Signal Frequency and Harmonics Estimation

Abstract

Frequency and Harmonics are two important parameters for power system control and protection, power system relaying, power quality monitoring, operation and control of electrical equipments. Some existing approaches of frequency and harmonics estimation are Fast Fourier Transform (FFT), Least Square (LS), Least Mean Square (LMS),
Recursive Least Square (RLS), Kalman Filtering (KF), Soft Computing Techniques such as Neural Networks and Genetic Algorithms etc. FFT based technique suffers from
leakage effect i.e. an effect in the frequency analysis of finite length signals and the performance is highly degraded while estimating inter-harmonics and sub-harmonics
including frequency deviations. Recursive estimation is not possible in case of LS. LMS provides poor estimation performance owing to its poor convergence rate as the
adaptation step-size is fixed. In case of RLS and KF, suitable initial choice of covariance matrix and gain leading to faster convergence on Mean Square Error is difficult. Initial choice of Weight vector and learning parameter affect the convergence characteristic of
neural estimator. Genetic based algorithms takes more time for convergence.