Robust frequency estimation in three-phase power systems using correntropy-based adaptive filter

• Published in: IET science, measurement & technology Vol. 9; no. 8; pp. 928 - 935
• Main Authors: Khalili, Azam, Rastegarnia, Amir, Sanei, Saeid
• Format: Journal Article
• Language: English
• Published: The Institution of Engineering and Technology 01.11.2015
• Subjects: Adaptive filters; Algorithms; augmented CLMS algorithm; augmented complex statistics; Clarke’s transformed three‐phase voltage; complex LMS algorithm; computational complexity; Computer simulation; correntropy‐based adaptive filter; Electric potential; frequency estimation; gradient methods; gradient‐ascent optimisation; harmonics suppression; higher order statistics; higher‐order harmonics; higher‐order moments; impulse noise; impulsive noise; linear predictive model; Mathematical models; maximum correntropy criterion; maximum entropy methods; model parameter estimation; Noise; power supply quality; power system parameter estimation; random noise; random noise source; robust adaptive algorithm; robust frequency estimation; Statistics; three‐phase power systems; unbalanced system conditions; Voltage; voltage sags; random noise source; complex LMS algorithm; augmented complex statistics; Clarke’s transformed three-phase voltage; model parameter estimation; correntropy-based adaptive filter; impulse noise; impulsive noise; unbalanced system conditions; gradient methods; adaptive filters; augmented CLMS algorithm; power system parameter estimation; linear predictive model; gradient-ascent optimisation; random noise; higher order statistics; robust adaptive algorithm; three-phase power systems; higher-order moments; frequency estimation; power supply quality; voltage sags; maximum correntropy criterion; higher-order harmonics; robust frequency estimation; maximum entropy methods; harmonics suppression; computational complexity
• ISSN: 1751-8822; 1751-8830
• DOI: 10.1049/iet-smt.2015.0018

In this study, the authors propose a robust adaptive algorithm for frequency estimation in three-phase power systems when the voltage readings are corrupted by random noise sources. The proposed algorithm employs the Clarke's transformed three-phase voltage (a complex signal) and augmented complex statistics to deal with both of balanced and unbalanced system conditions. To derive the algorithm, a widely linear predictive model is assumed for the Clarke's transformed signal where the frequency of system is related to the parameters of this model. To estimate the model parameters with noisy voltage reading, they utilise the notions of maximum correntropy criterion and gradient-ascent optimisation. The proposed algorithm has the computational complexity of the popular complex least-mean-squares (CLMS) algorithm, along with the robustness that is obtained by using higher-order moments beyond just second-order moments. They compare the performance of the proposed algorithm with a recently introduced augmented CLMS (ACLMS) algorithm in different conditions, including the voltage sags and presence of impulsive noises and and higher-order harmonics. Their simulation results demonstrate that the proposed algorithm provides improved frequency estimation performance compared with ACLMS especially when the measured voltages are corrupted by impulsive noise.