Arc Diagnosis Method Of Series Fault In Photovoltaic System Under Line Impedance Interference

The period of the "14th Five-Year Plan" is a critical time for the implementation of the carbon peaking and carbon neutrality strategy.The proportion of clean energy generation such as photovoltaics,hydropower,and wind power in the power system continues to increase,and the overall energy structure is continuously optimized.However,due to exposure to a constantly changing environment,DC lines may damage the insulation of wires due to rust and corrosion,or cause wire breakage,resulting in the occurrence of arc faults.The operating conditions of the photovoltaic DC system are complex,and the operating characteristics of household appliances can also suppress the arc fault characteristics during the detection process.In addition,the impedance of the line weakens the characteristic values,making arc fault characteristic extraction a difficult problem.This paper takes the series arc fault in the photovoltaic system as the research object,considering the weakening of characteristic values and noise interference caused by line impedance,proposes a high-precision series arc fault detection and positioning algorithm,and designs a arc fault real-time diagnosis system based on the Raspberry Pi.By conducting in-depth research and modeling analysis on the Cassie arc model and photovoltaic cell model,a photovoltaic system arc fault model was established based on these two models.By tuning the arc fault model,line impedance,and load parameters,current waveforms were collected to obtain the simple characteristics of the current signal.It was pointed out that when a arc fault occurs,the circuit current decreases and exhibits significant fluctuations.The simulated current waveform was then compared with the experimental waveform to verify the correctness of the simulation experiment.Finally,wavelet analysis was performed on the circuit current to better reflect the changes in frequency spectrum amplitude after a arc fault occurs.Based on the relevant standards,a photovoltaic system series arc fault experimental platform was constructed.Series arc fault experiments with pure resistive load and inverter load were conducted under different line impedances,and corresponding arc fault current and voltage waveforms were collected using sensors to study the characteristics of photovoltaic system series DC arc faults.Then,time-frequency domain analysis was performed on the current signal,and the wavelet transform method was introduced.On this basis,the collected current signal was compared and analyzed,and the Rbio3.1 wavelet basis,6-layer decomposition,and 15.625-23.438 k Hz feature frequency band were selected by comparing the enhancement ratio.It was found that the feature of the current signal significantly increases in the fault state,which can distinguish between normal and fault states.Moreover,the feature value is closely related to the location of the series DC arc fault,and increases as the cable length decreases.Therefore,the Rbio3.1 wavelet basis was chosen as the input feature quantity for later arc fault detection and location methods.A LightGBM detection and location model was established,and the samples were divided into a learning set and a testing set.The model was optimized by particle swarm algorithm to select various parameters and solve the problem of hyperparameter optimization of LightGBM.The detection and location model was trained,and the trained model was used for location detection and error analysis.The results showed that the accuracy of the constructed model in detection and location was 93.2%,which can meet practical requirements.The detection and positioning device was designed,the software running environment was built,and the overall process was designed.The trained LightGBM model was transplanted into a detection and positioning device to complete real-time detection and location experiments,further expanding its application.In this paper,a detection and location program was developed based on the typical structure of a photovoltaic system series arc fault model,the detection and location method of a photovoltaic system series arc fault was verified,and the detection and positioning device was designed.The results showed that the feature extraction of the photovoltaic system series arc fault based on wavelet transform and the detection and location method of the photovoltaic system series arc fault based on the LightGBM algorithm are effective and feasible.