Research On Methods Of Harmonic Measurement And Harmonic Source State Recognition In Power Grid

With the development of Industry 4.0,new energy power generation,electric vehicles and high-speed railways,the problem of harmonic pollution in the power grid has become more and more serious,and the comprehensive management,division of responsibilities and cause analysis of the power grid harmonics have become urgent.The research on power grid harmonic measurement and harmonic source status identification is a prerequisite for the comprehensive management,responsibility division and cause analysis of power grid harmonics,and it is of great significance.Therefore,this paper uses the main line to study the methods of power grid harmonic measurement,power grid harmonic source location identification and power grid harmonic source type identification respectively,and use virtual instrument technology to realize the main content as follows:(1)Analyzed the basic principles of Fast Fourier Transform(FFT)and its limitations in measuring steady-state harmonics,and suppressed the effect of spectrum leakage based on the main lobe performance and side lobe performance of different window functions,Constructed a third-order RVB hybrid self-convolution window with Rife-Vicent window and Blackman window as the base window,its sidelobe performance is better than the base window function,and deduced the phase difference amplitude of the third-order RVB hybrid self-convolution window The frequency correction formula has a significant effect on weakening the fence effect.The improved Hilbert Huang Transform(HHT)is introduced to measure transient harmonics,time-frequency analysis and parameter measurement of transient harmonic start and end time.The simulation results show that the third-order RVB hybrid self-convolution window phase difference correction FFT has lower relative error in measuring steady-state harmonic parameters than traditional window function phase difference correction FFT,and has better anti-interference and robustness.Improved HHT can better perform time-frequency analysis of transient harmonics.The relative error of the parameters at the start and end of the measurement is relatively low and it is not sensitive to noise.(2)Based on the harmonic voltage and current data measured at the common point,the equations are established using Partial Least Squares Regression(PLSR),and the regression equations considering the user-side impedance and not considering the user-side impedance are respectively derived.Intercept coefficients and regression coefficients can identify the location of harmonic sources,but they are susceptible to abnormal grid disturbances and harmonic fluctuations,which will cause the estimated value to deviate from the original value.Therefore,this paper first uses Mahalanobis distance method to filter the abnormal disturbance data,secondly uses the box plot method to filter the harmonic fluctuation data,and then uses the filtered data for PLSR,which can reduce the influence of abnormal disturbance and harmonic fluctuation to Lower level.The simulation results show that,regardless of whether the user-side impedance is considered,the screening effect is better than the unscreened effect,and the location of the harmonic source can be accurately identified.(3)After the harmonic source is measured and the position is identified,the relevant data obtained are synthesized,and the amplitude-frequency statistical characteristics including the center of gravity frequency and frequency standard deviation are further extracted,and the random forest method with both interpretability and accuracy is used to compare the harmonics.The types of wave sources include six-pulse rectification harmonic sources for identification.It focuses on analyzing the relevant knowledge of decision tree and its integration method,comprehensively evaluating the accuracy rate and recall rate,etc.to evaluate the classification model,extracting and processing features,and generating the random forest model.The simulation results show that the F1 value of the random forest method for identifying the electric arc furnace type harmonic source type is as high as 94.37%,and the F1 value for identifying the twelve-pulse rectification type harmonic source type is as low as 92.09%,and the total recognition accuracy rate can reach 92.68%,the recognition effect of the harmonic source type is good.Finally,based on the method of this article and virtual instrument technology,the power grid harmonic measurement and harmonic source state recognition instrument is developed,and the composition of the instrument hardware platform is introduced.The software module design is carried out on the basis of the existing instrument hardware platform,and the core computing part is used Mixed programming implementation.