Title: Assisted decision making for conditional maintenance of distribution network equipment based on multi-task deep learning

Authors: Zhenyu Luo; Tu Xiong; Minghui Chen; Mingzhu Kong; Chunkai Zhang

Addresses: Guangzhou Power Supply Bureau, Guangdong Power Grid Co., Ltd., Guangzhou, Yuexiu District, China ' Guangzhou Power Supply Bureau, Guangdong Power Grid Co., Ltd., Guangzhou, Yuexiu District, China ' Guangzhou Power Supply Bureau, Guangdong Power Grid Co., Ltd., Guangzhou, Yuexiu District, China ' Dongfang Electronics Corporation, Yantai, Shandong, China ' Dongfang Electronics Corporation, Yantai, Shandong, China

Abstract: In order to solve the problems of low accuracy, long time and high cost in traditional decision-making methods, an assisted decision-making method for conditional maintenance of distribution network equipment based on multi-task deep learning is proposed. Install sensors on key equipment in the distribution network, combine with decision trees to obtain preliminary mining results of distribution network equipment status information, and cluster and fill in the preliminary mining information. The processed information is input into a multi-task deep learning network, and the auxiliary decision-making results for distribution network equipment status maintenance are obtained through the operations of the representation layer, distribution network equipment operation status discrimination layer, maintenance period prediction layer and multi-task learning layer. The experimental results show that the maximum decision accuracy of the proposed method is 97.6%, the decision time is always below 74 ms and the total maintenance cost is 1.979 × 10⁵ yuan.

Keywords: multi-task deep learning; distribution network equipment; conditional maintenance; assisted decision making; decision trees; cluster; fill.

DOI: 10.1504/IJCAT.2024.141362

International Journal of Computer Applications in Technology, 2024 Vol.74 No.1/2, pp.63 - 72

Received: 09 Oct 2023
Accepted: 13 Feb 2024
Published online: 09 Sep 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article