Research On Harmonic Detection Method Of Power System Based On Local Mean Decomposition

With the grid connection of distributed new energy and more and more nonlinear loads are used in the power system,harmonic pollution has become more serious,and harmonics can have catastrophic consequences for the grid and other loads,such as system resonance,equipment failure and communication interference,etc.Accurate detection of the frequency,amplitude and phase of harmonics is of great significance for harmonic control in power systems.In recent years,Local Mean Decomposition(LMD)has played an increasingly important role in the field of power system harmonic detection,but its own modal aliasing and endpoint effects have seriously affected its application value.These two problems are improved respectively,and the improved algorithm is applied to the field of harmonic detection.This paper will work on the following aspects:1.By comparing the local mean decomposition with the widely used empirical mode decomposition(EMD),it is concluded that LMD is superior to EMD in avoiding false components.By analyzing the advantages and disadvantages of LMD through simulation experiments,it is concluded that the endpoint effect and modal aliasing still affect its application in the field of harmonic detection,which provides theoretical support for subsequent chapters.2.Compared with the traditional harmonic detection method,Prony method has the advantages of high resolution and simple calculation,but its anti-noise performance is poor.Aiming at the disadvantage that Prony is susceptible to noise when identifying harmonic parameters,LMD is applied to the noise reduction of harmonic detection,and the modal mixing effect of traditional LMD will affect the denoising effect.Firstly,the ensemble local mean decomposition(ELMD)method for modal aliasing optimization is used to decompose the harmonic signal to obtain a series of PF components.Then,the Kullback-Leibler Divergence(K-L)is used to distinguish the noise component and the effective component,and the effective component is denoised by singular value decomposition(SVD).On this basis,combined with Prony method,a harmonic detection method based on ELMD-SVD and Prony is proposed.Finally,the harmonic signal of power system is simulated and compared with other commonly used harmonic detection methods.By analyzing the reconstructed waveform and signal-to-noise ratio,root mean square error data and harmonic detection accuracy,it can be seen that the reconstructed waveform of ELMD-SVD-Prony method is more consistent with the original signal,and the signal-to-noise ratio is improved more,the root mean square error is lower,and the harmonic detection accuracy is higher.3.BP neural network harmonic detection has the advantages of high precision and simple and reliable,but the training of BP neural network needs to predict the number of main frequencies and frequency parameters in advance.To solve this problem,the LMD is combined with the BP neural network for the first time.The LMD is used to detect the number of main frequencies of the signal and extract the frequency parameters to provide them to the BP neural network to solve the parameter setting problem of the BP neural network.But to accurately detect the harmonic frequency,it is necessary to suppress the LMD endpoint effect.Firstly,aiming at the problem that the endpoint effect affects the number of harmonic main frequencies and frequency parameters extracted by LMD,the signal extension is carried out by boundary local feature scale extension(BLFSE)method,and it is adaptively embedded in LMD,and Boundary Local Mean Decomposition(BLMD)algorithm is proposed to suppress the endpoint effect of traditional LMD.Then the BLMD-BP neural network harmonic detection method is proposed based on BP neural network.The number and frequency parameters of the main frequency given by BLMD are input into BP neural network,and then more accurate frequency,amplitude and phase data are obtained by BP neural network.Finally,the BLMD-BP and FFT(Fast Fourier Transform)-BP are simulated and compared,and the results show that the BLMD-BP has higher harmonic detection accuracy.4.The proposed two methods are used for harmonic detection of three power system signal simulation models,one experimental data and three actual harmonic current signals of nonlinear loads.The results show that ELMD-SVD-Prony consumes longer machine time,but has good adaptability to both stationary and non-stationary signals,and the denoising effect is better.SVD-BLMD-BP method has weak adaptability to non-stationary signals,but it takes shorter time and has higher detection accuracy for stationary signals than the former.The current signal detection results of nonlinear load show that the two methods can effectively identify the actual harmonic signal and have strong practical value.Different algorithms can be selected according to the power system online / offline power flow calculation requirements.