Researches On Harmonic Detection Algorithms Based On Convolution Windows And Revised DFT

In recent years,the proportion of various harmonics in power grid rises sharply,due to the quantity of nonlinear elements increasing and distributed energy resources(DER)taken into application.To estimate harmonic parameters precisely and to address the questions caused by harmonics,four aspects of existing windowed and interpolated DFT(IpDFT)methods are revised in this paper.In the aspect of de-noise: Algorithms including wavelet decompositions in time-and-frequency domain and Empirical Mode Decomposition(EMD)are simulated and analyzed.The rising degree of the signal-to-noise ratio(SNR)of the signal to be detected is large in a low-SNR environment,however it performs not eminent in a high-SNR environment.To overcome this weakness,a de-noise method based on standard unit period(SuP)is proposed.It utilizes the periodicity character of harmonics,and constructs weight-vector based on 3-sigma rule.Based on the correction,the SNR could be improved remarkably.Meanwhile,the effective components would not be filtered.In the aspect of normal harmonics detection: IpDFT methods based on Rife-Vincent window are simulated,and it's found that if the window contains lower sidelobes and a narrower mainlobe,the errors caused by spectral leakage become less.Therefore,the 4th-order Hann self-convolution window is applied to restrain the errors from spectral leakage.Meanwhile,merely two spectra are needed to construct interpolation equation due to the brief expression of Hann window in the frequency domain,and this equation can be solved analytically,which lead to the errors caused by picket-fence effect hardly remain.In the aspect of inter-harmonics detection: The reasons to producing mainlobe interference(MLI)are analyzed,and inter-harmonics are divided into 3 types.For the two types accompanying with MLI,low interference zone(LIZ)is partitioned based on a 2nd-order Hann and 6th-order Rect.window(H2R6 window),then the parameter values of can be estimated.Considering this method is not universal,another method based on a 5th-order Tria.and 11th-order Rect.window(T5R11 window)and Accelerated PSO is proposed.The solution of pre-calculation could be optimized based on the discrepancy between practical spectra and the spectra reconstructed.Additionally,it is considered that methods like PSO cost a lot of time.Thus,the proposed PSO is accelerated in three approaches including locking the frequency,using changing inertia weight and utilizing the T5R11 window.These revisions aim to rise the parameters estimation accuracy and keep the algorithm able to be executed real-timely.Finally,a 14-bus network with distributed generators(DG)is constructed in order to verify the proposed method can be applied in practically.The solar arrays and the wind turbines supply power for this system,and arc furnace serves as the typical nonlinear load,to produce various harmonics and inter-harmonics.The results show that each electrical signal can be simulated well.Meanwhile,the detection accuracy and in-time capability are all meet national standards and international standards.