Research On A Fault Source Location Method Of Power Grid Voltage Sag Based On Random Matrix Theory

With the continuous development of intelligent power grids,the network structure of power grids has become more complex,and numerous refined and sensitive devices are used in large quantities in power grids,the impact caused by voltage sags is gradually receiving wide attention in the power field.In the grid,system short-circuit fault is the main cause of voltage sag problem.How to remove the fault as soon as possible after the fault occurs to restore power supply is particularly important.Accurately locating the source of voltage sag fault can effectively shorten the restoration time and solve the differences between power supply department and power users regarding the cause of voltage sag.The thesis focuses on the following aspects of the key issues in locating the source of voltage sag faults in power grids.(1)Considering the problem that most traditional methods of voltage sag fault source location are affected by the accuracy of the physical model established during the location process,a random matrix-based voltage sag fault source location model is established by taking advantage of the fact that random matrix theory does not rely on the simplification and assumption of the system model and can extract effective information from a large amount of data.Firstly,the influencing factor data and the grid operation state data are augmentation matrix.Then,the mean spectral radius of the characteristic statistic is used to construct an index to analyze the correlation size between each influencing factor and the grid state to realize the voltage sag fault source location.Finally,the proposed method is verified to be effective in locating the source of voltage sag faults through the analysis of arithmetic cases.(2)In complex voltage sag events,multiple voltage sag events overlap together,adding difficulties to the accurate location of the voltage sag fault source.A fault source location method in complex voltage sag events that combines random matrix theory and the traditional disturbance power and disturbance energy method is proposed.The method first uses the random matrix theory-based voltage sag fault source location model to determine the suspicious area of the voltage sag fault source.Then,the atomic algorithm is used to divide the disturbance signals of each node in the suspicious area into time series zones,and the disturbance active power and disturbance energy methods are combined to locate and analyze the time series zones of each node separately,which can accurately locate the location of each fault source in the complex voltage sag.Finally,the proposed method is verified to be able to effectively locate each time source in the complex voltage sag event through the analysis of arithmetic cases.(3)In complex power networks,due to the complex network structure and the large amount of collected data,most existing localization methods need to traverse all possible fault points,resulting in long computing time.To address this problem,this thesis applies the community structure theory to partition the complex electric network,and then uses the random matrix theory-based voltage sag fault source location model to locate each partition and realize the fault source location according to the analysis results.Finally,the results show that the method is effective and feasible through simulation.(4)In the process of grid partitioning,it leads to the problem that the number of state data dimensions is too low or the networked partitioning is too much,and the mean spectral radius does not apply to all dimensions of the matrix.To deal with this problem,a voltage sag fault source location method based on random matrix theory and convolutional neural network is proposed.Firstly,the grid historical data are analyzed by random matrix theory to obtain four kinds of characteristic statistics data: mean spectral radius,standard deviation of eigenvalue,maximum eigenvalue and minimum eigenvalue in different partitions and different fault cases.Then,convolutional neural networks are used to train these statistics data and extract the optimized characteristic statistics for the voltage sag fault source location model.Finally,the effectiveness and accuracy of the method are verified by arithmetic simulation and comparative analysis.