Research On Fault Location Method Of Power System Based On Pmu Missing Data Reconstruction

As the scale of power system expands,the system for controlling and protection has become more and more complex,and the extent of fault’s impact grows.As a result,enhancing the fault location capability of the power system has become one of the most important factors in ensuring its stable operation.The Phase Measurement Unit（PMU）is a real-time,synchronicity,accuracy and comprehensive power system measurement and acquisition device that addresses the shortcomings of traditional power system measurement and acquisition devices,such as long measurement and acquisition cycles,asynchronous measurements and weak dynamic characteristics.However,PMU data may be missing owing to communication congestion,hardware failure or transmission delay.It has a direct impact on the power system’s online monitoring and fault location,resulting in incalculable economic damages.Based on the analysis of the current situation of PMU product application,the research of PMU missing data reconstruction and the research of power system fault location,this thesis proposed a fault location method for power system based on PMU missing data reconstruction.For optimal PMU configuration,an IEEE 39 bus power system model was developed.Generative adversarial network（GAN）was utilized for PMU missing data reconstruction,and a two-step method including the phase difference search method and the impedance method was used to achieve fault location in power system.The research’s specifics are as follows:（1）The optimal configuration method of PMU under the constraints of global viewability was proposed using the IEEE 39 bus power system model.The crossover and mutation probability were enhanced using Genetic Algorithm,and the optimal goal was to reduce the number of PMU configurations to achieve the global monitoring of the power system.It provided a foundation for reconstructing PMU missing data and locating power system’s fault.（2）To overcome the communication congestion and missing data of PMU,a missing data reconstruction based on GAN was proposed.For data reconstruction,GAN was employed in this method to actively learn the distribution patterns among PMU data in the power system through an unsupervised form.When the weights were updated in training process of GAN,a high-performance stochastic optimization gradient descent algorithm was introduced when the weights were updated to improve the training speed.（3）Based on the reconstruction of the missing data of the PMU,the phase difference search method was used to determine the fault occurrence region,and the impedance method was used to determine the fault occurrence point for the region where the fault occurs.According to the above,a two-step method based on the power system fault location method was constructed.Using the IEEE 39 bus power system model as the simulation object,it was discovered that the relative error was not higher than 5% and the time was less than 5s,which demonstrated the effectiveness of the proposed two-step method based on power system fault location method.