Open Access

ARTICLE

A Missing Data Complement Method Based on 3D Convolutional Neural Network and CGAN for a Distribution Network

Kewen Li, Xiaoyong Yu, Shifeng Ou^*, Jueming Pan
Electric Power Science Research Institute of Guangxi Power Grid Co., Ltd., Nanning, 530023, China
* Corresponding Author: Shifeng Ou. Email:
(This article belongs to the Special Issue: Advanced Analytics on Energy Systems)

Energy Engineering https://doi.org/10.32604/ee.2025.073825

Received 26 September 2025; Accepted 25 November 2025; Published online 30 January 2026

Abstract

The increasing integration of renewable energy sources (e.g., wind and solar power) into distribution grids and the development of new, source–grid–load–storage coordinated power systems have led to a substantial expansion in the volume of situational awareness data in the distribution networks. Moreover, the transmission of low-voltage distribution measurement data via a power line carrier (PLC) is often susceptible to packet loss and, consequently, data gaps. To address these issues, this paper proposes a data completion method using a conditional generative adversarial network (CGAN) integrated with a three-dimensional convolutional neural network (3D-CNN). This approach leverages the ability of CNNs to extract and fuse multidimensional spatiotemporal features and the power of GANs (generative adversarial networks) for data augmentation. Firstly, a 3D-CNN is trained to establish a mapping between the spatiotemporal context of the measured data and the target missing data. Secondly, a CGAN is practicing via adversarial training to establish a data completion model for the distribution networks. Finally, the simulations of the IEEE 14-bus and 33-bus systems demonstrate the proposed approach’s performance improvement in the distribution networks compared with that of conventional methods in terms of root mean square error, spatiotemporal correlation, and maximum volatility amplitude, which are the typical measuring metrics.

---

Keywords

Distribution network; data estimation; 3D convolutional neural network; conditional generative adversarial networks; spatiotemporal correlation

---