Fault zone diagnosis of three-terminal hybrid UHVDC transmission lines based on multi-mode decomposition and multi-branch parallel residual network

The randomness and non-linearity of fault electrical signals variations in the event of earth fault on a three-terminal hybrid ultra high voltage direct current（UHVDC） transmission line. The regularity of fault electrical signals is even weaker under noise interference，making it difficult to extract fault characteristics quickly and accurately and therefore making diagnosis of fault areas difficult. As a result，a diagnosis model that integrates current waveform feature extraction and fault zone diagnosis is proposed. The characteristics of the current waveforms during faults in different areas are analyzed. The fault currents are decomposed by using three algorithms with different mathematical mechanisms，which prevents some modal subsequences that are prone to confounding from escaping in a single decomposition method. The multi-branch parallel residual network is then used to rapidly mine the multi-scale spatially coordinated interactive features of the decomposition components，in combination with the gated recurrent units module，to further extract the temporal characteristics of the fault currents，which enhances the deep features significantly. The key parameters in this model are optimized with sparrow search algorithm. By applying these steps，a network structure fully adapted to power grid fault diagnosis is constructed and the fault areas are quickly diagnosed. The simulative results show that the scheme has high sensitivity，strong anti-interference ability，and meets the requirements of reliability and quickness. © 2024 Electric Power Automation Equipment Press. All rights reserved.