To solve the problem of difficult identification of the internal overvoltage category of the distribution network, enhance the mastery of the internal voltage identification technology of the distribution network, a method of identifying the internal overvoltage of the distribution network based on the convolutional neural network is proposed. The effectiveness and superiority of the proposed evaluation method are illustrated by an example analysis of a medium-voltage distribution network in East China demonstrates the effectiveness and superiority of the proposed evaluation method. The results show that under ten tests, ACC1 and ACC2 of each type of overvoltage are tested. In the cross-validation process, only two cases of misjudgments occur when switching capacitor banks, while the recognition rates of the remaining six types of overvoltage are all at 100%. The total recognition rate of the test samples is 99.57%, indicating that the features extracted by convolutional neural network (CNN) have covered all the seven types of overvoltage. The CNN network constructed by the proposed method recognizes the 110 overvoltage samples obtained, and its recognition accuracy reaches 100%. It shows that the proposed recognition method has a high recognition rate and verifies the effectiveness and superiority of CNN.
Cite this article as: F. Long, H. Xu, W. Zhan, Y. Wang, C. Zou and J. Bhola, "Recognition of internal overvoltage in distribution network based on convolutional neural network," Electrica., 22(3), 342-350, 2022.