Algorithm Research And System Design Of Fault Diagnosis And Location Based On PMU For Distribution Network

With the overall construction of China’s electricity internet,distributed power generation has been increasingly promoted.Many industrial terminals,such as electric vehicles,are connected to the distribution network,and the reliability of power supply in the distribution network faces serious challenges.For a long time,the development of fault detection technology for distribution networks has been slow,and traditional fault diagnosis and location algorithms are prone to interference from network analysis or external factors,resulting in abnormal results and a waste of human and material resources.As the main structure of the wide-area measurement system,the synchronous phasor measurement unit(PMU)can obtain time-calibrated measurement data dynamically and quickly by using the global positioning system,and has broad application prospects in power system fault detection.However,there are corresponding configuration difficulties in the development of PMU-based accurate fault location technology for medium voltage distribution networks: How to achieve low-cost rapid fault detection based on limited PMU nodes is still an urgent issue,and the correct response to weak faults needs to be combined with emerging technologies.In response to the difficulties and challenges mentioned above and the actual needs in China,this paper proposes to introduce a fault diagnosis and location system based on synchronous phasor measurement devices(PMU)in distribution networks,thereby achieving a breakthrough in the performance of fault diagnosis and location technology in this field.The main work of this paper is as follows.(1)With the relevant theories of PMU,a fault data feature establishment method based on PMU collection is designed.This method first extracts waveform information such as voltage and current from the power grid through PMU,extracts IMF components based on set empirical mode decomposition,and performs normalization and filtering.Then,it extracts energy spectral density features based on wavelet transform,further enhancing the ability of fault feature energy to express power grid status.Finally,it uses principal component analysis to reduce data dimensions and reduce the computational pressure of subsequent algorithms.(2)Based on the fault characteristics achieved by the above methods,a fault diagnosis and location method for distribution networks based on fuzzy reasoning and deep learning model is studied and designed.This method uses a fuzzy Petri net algorithm for fast and scalable fault diagnosis.After identifying the type of fault,a deep neural network based on residual blocks is used for fault location to achieve accurate measurement of fault distance.(3)Based on the theoretical methods of the above research,the hardware architecture and application software design of a fault diagnosis and location system based on PMU nodes are completed,including a data acquisition unit based on PMU(SMU-2 type),a signal processing unit based on TMS320C6713,and a management unit based on STM32F103CBT6.The simulation experiment results show that the algorithm and system designed in this paper have high accuracy and practicality: the diagnostic accuracy rate for ground fault is above 96.4%,the diagnostic accuracy rate for short circuit fault is above 96.8%,and the overall fault location error of the system is within 100 m.