UHVDC Transmission Line Fault Identification Method Based on Generalized Regression Neural Network

A protection method for ultra-high voltage direct current transmission lines based on generalized regression neural network (GRNN) is proposed to address the issues of easy rejection and long fault detection time in ultra-high voltage direct current protection. Firstly, based on the generalized S-transform, the fault characteristic information in the frequency domain is obtained to construct the input data for GRNN. Secondly, the chaos quantum particle swarm optimization (CQPSO) algorithm is used to optimize the parameters of the generalized regression neural network, form an ideal network model based on the principle of the lowest fitness function, and better learn the fault characteristics of the ultra-high voltage DC transmission line. The Softmax classifier is utilized to classify deep-level features, identifying faults as external, bus, or line faults, and polarizing them into positive, negative, or bipolar faults, then outputting recognition results. Finally, the ultra-high voltage direct current transmission model built in the PSCAD/ EMTDC simulation environment is validated, and the validation results showed that the proposed method has good performance in fault detection and fault pole selection of ultra-high voltage direct current transmission line relay protection. Compared to traditional convolutional neural networks, generalized regression neural networks, support vector machines, and other methods, the fault recognition accuracy of the proposed method in this paper has been improved by 6. 6% , 0. 65% , and 7. 69% , respectively, meeting the requirements of protection speed and reliability.