期刊
IEEE ACCESS
卷 9, 期 -, 页码 158961-158971出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/ACCESS.2021.3129780
关键词
Training; Neurons; Power grids; Dispatching; Data models; Data mining; Visualization; Deep belief network; faulty feeder detection; power dispatching system data; single-phase ground fault
资金
- National Natural Science Foundation of China [51967002]
- Science and Technology Projects of China Southern Power Grid Company Ltd. [GXKJXM20190619, GXKJXM20190680]
This paper proposes a faulty feeder detection method based on Deep Belief Network for neutral non-effectively grounded systems. The method achieves a high accuracy of 94.7% by using millisecond-level data directly from the power dispatching system for training.
This paper proposes a faulty feeder detection method based on Deep Belief Network (DBN) of deep learning theory for the neutral non-effectively grounded systems. It consists of two steps: firstly, a DBN-based faulty feeder detection model is built with feeder current, power and power factor as input feature parameters. Then, the input feature data are obtained during the single-phase ground fault from the master station of power dispatching system, which will construct a training set. By unsupervised pre-training and supervised fine-tuning, the proposed model obtains the mapping relationship between raw data and fault characteristics and realizes the faulty feeder detection. The advantage of the proposed method is using millisecond-level data in power dispatching system directly. Moreover, the sampling device does not need to install, which significantly reduces the construction costs and is of strong adaptability. The analyzed result using the ground fault data of an actual substation for more than two years shows that the proposed method has a better performance than SVM and BP neural network, and the accuracy is up to 94.7%. The proposed method has been implemented in Lipu Power Grid, Guangxi, China with excellent application effect and extensive application prospects.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据