Research On Micro-Grid Fault Detection Based On Deep Learning

With the rapid development of renewable energy power generation,the microgrid has attracted the attention of the industry because of its advantages of suppressing power fluctuations caused by the access of a large number of distributed power sources and improving the reliability of equipment power supply.However,during the operation of the microgrid,due to weather factors and other reasons,the circuit will have open circuits,short circuits and arc faults that affect the normal power supply.In severe cases,an electrical fire may even cause damage to the entire power grid and cause cascading failures.Therefore,it is of great significance to study the microgrid fault detection technology and improve the detection accuracy for the safe operation of the microgrid.In order to improve the accuracy of photovoltaic microgrid fault detection,this paper builds a photovoltaic microgrid system and studies the photovoltaic microgrid fault detection methods under various faults.The main contents are as follows:(1)On the basis of understanding the structure of each component of the photovoltaic microgrid system,this paper first analyzes the basic structure of the photovoltaic cell and builds the corresponding mathematical model;second,it analyzes the principle of the battery energy storage system and builds a simulation model;finally,Design the AC microgrid and simulate the other components.(2)In the photovoltaic micro-grid system,the cable line usually has open circuit and short circuit faults due to external forces such as external animals eating and eating during normal operation.Using signal time-frequency analysis method to decompose the signal and construct a classifier is an important solution to the fault detection of the microgrid.This paper analyzes the disadvantages of component waveform distortion and false components in the traditional intrinsic time scale decomposition method,studies the effect of using different interpolation methods on curve fitting,and proposes an improved intrinsic time scale decomposition method,Referred to as MITD).The simulation results show that,compared with the traditional ITD,the proposed method can both smooth the signal and retain the characteristic data effectively.In order to solve the problem of interruption and short circuit fault classification of photovoltaic microgrid,this paper designs a MITD-DNN joint multi-feature fusion fault detection scheme.The fault detection scheme first collects the three-phase voltage amplitude of the branch,decomposes the collected data using improved ITD to extract feature values,and after feature selection,the fused feature matrix is put into a deep neural network(DNN)for training,and According to the classification results of the deep neural network,the fault type and fault phase are accurately discriminated.In order to evaluate the performance of the proposed fault detection scheme,a comprehensive evaluation study was carried out on the microgrid built based on the IEC-61850 standard.The experimental results show that the detection scheme can effectively distinguish open circuit and short circuit under different photovoltaic microgrid parameter configurations Fault conditions.(3)In the photovoltaic micro-grid system,due to the great influence of the weather,especially the thunderstorm weather will cause the arc fault of the load.Therefore,in view of the occurrence of noise in the actual measured waveform data due to equipment errors and other reasons when an arc fault occurs in the microgrid,a detection method combining EEMD denoising and MITD decomposition is proposed.First,use EEMD to decompose the load current waveform with noise added,and use the correlation coefficient to remove the false component to reconstruct the current signal,and then select the relevant features according to the variance selection method.Second,use the MITD to decompose and extract the reconstructed current signal Feature,and finally merge the selected feature with the original current data into a feature matrix and input it into CNN for classification.Through experimental verification,the method proposed in this paper can effectively detect arc faults and ensure the safe and stable operation of photovoltaic microgrid.