Research on Overvoltage Hierarchical Identification Method Based on Modular Hierarchy and Salient Feature Quantity

During overvoltage faults in systems with a neutral point that is not grounded, the complexity of overvoltage characteristic quantities and the vast amount of data make it challenging to accurately identify overvoltage waveforms. If a unified identification approach is taken for all overvoltage characteristic quantities, the unclear structural hierarchy can lead to a higher likelihood of misjudgment, a massive workload, poor real-time diagnostic capabilities, and low efficiency, making online judgment difficult. In this paper, a classification structured tree is constructed to establish multiple modular levels based on the principle that the similarity of different types of overvoltages increases from low to high. By analyzing the characteristic quantities of three-phase voltage and zero-sequence voltage in the time domain, frequency domain, and wavelet energy spectrum, prominent characteristic quantities for specific types of overvoltage are set at each level. Fuzzy clustering based on genetic algorithms is then employed to analyze the overvoltage signals at each level, enabling layer-by-layer clustering, identification, and multi-level classification filtering of various types of overvoltage.