基于深度信念网络的直流输电线路换相失败预测技术

打开文本图片集

中图分类号：TB9；TM73 文献标志码：A文章编号：1674-5124（2025）08-0155-08

Abstract： In response to the large-scale integration of new energy into the power grid， the randomness of commutation failure in DC transmission systems becomes stronger，and the influencing factors become more complex，leading to increased diffculty in predicting commutation failure.A DC transmission line commutation failure prediction technology based on deep belief networks isproposed. Firstly， the characteristics ofcommutationfailure areanalyzed，and theACvoltage，DCcurrent，and trigger angle fault data on the inverter side are extracted. Secondly， by utilizing the advantage of deep belief networks in learning high-dimensional data with few unsupervised features，the fault data is normalized as input data for deep belief networks，and a DC transmisson line commutation failure prediction model is constructed. Finally， the Softmax classifier outputs a commutation failure label to achieve commutation failure prediction.A PSCAD/EMTDC DC transmission model is built for verification， and the experimental results show that the proposed method has high accuracy in predicting commutation failure. Compared with common convolutional neural networks and extreme learning machines， the proposed method has improved by 10.6% and 8.5% 0 respectively， verifying its effectiveness.

Keywords： new energy; deep belief network; DC transmisson lines; commutation failure; Softmax classifier

0 引言

随着新型电力系统建设的推进，分布式新能源接入电网容量不断提升，高压直流输电技术能够有效消纳多余的可再生能源[1-2]。（剩余13356字）