Abstract: The photovoltaic (PV) grid is affected by weather factors and nonlinear loads, and the disturbance components in the direct current (DC) signal make it difficult to ensure the accuracy of power quality assessment. Therefore, in this paper the problem that the composite multiscale fuzzy entropy (CMFE) can overcome the sudden change of the initial single component similarity measure of the photovoltai DC signal is utilized, then the regularized CMFE operator is constructed to evaluate the complexity of each initial single component after reconstruction, while constraining the residual energy to be minimized, and finally the separation of electrical signals and noise and other disturbance is realized. On this basis, an adaptive denoising method for photovoltai DC signal based on partial reconstruction of symplectic geometry mode decomposition (PRSGMD) is proposed, and combined with the advantage that extreme gradient boosting (XGBoost) can effectively mine the relationship between features and transient stability, the separation and identification of compound disturbance in photovoltaic DC signals is realized.

    Key words: photovoltaic; power quality disturbance identification; partial reconstruction of symplectic geometry mode decomposition;extreme gradient boosting