Abstract: An arc fault detection method for a photovoltaic system based on an adaptive kernel function and instantaneous frequency estimation includes steps of: sampling signal xt in a time window length of TNCT and obtaining an iterative time-frequency diagram of xt by nonlinear chirplet transform; extracting detection variables based on frequency component in the selected iterative time-frequency diagram to determine a moment when spectrum energy increases; when the moment is found, obtaining a matrix distribution form of the xt in time-frequency domain obtained by the adaptive optimum kernel time-frequency representation, and processing the matrix with sum of squares in a time dimension to obtain a column vector; processing each selected frequency bands with integration operation in a frequency dimension to obtain multiple detection variable values as inputs of a well-trained Naive Bayes model.

Abstract: An arc fault detection method for a photovoltaic system based on an adaptive kernel function and instantaneous frequency estimation includes steps of: sampling signal xt in a time window length of TNCT and obtaining an iterative time-frequency diagram of xt by nonlinear chirplet transform; extracting detection variables based on frequency component in the selected iterative time-frequency diagram to determine a moment when spectrum energy increases; when the moment is found, obtaining a matrix distribution form of the xt in a time-frequency domain obtained by the adaptive optimum kernel time-frequency representation, and processing the matrix with sum of squares in a time dimension to obtain a column vector; processing each selected frequency bands with integration operation in a frequency dimension to obtain multiple detection variable values as inputs of a well-trained Naive Bayes model; and determining photovoltaic system states in a current period.