Journal of Engineering, Project, and Production Management, 2026, 16(4), 2025-237
Multi-Scale Convolutional VAE for Anomaly Detection in Sensitive Power Marketing Data
1
Specialist, Marketing Service Center, State Grid Shanxi Electric Power
Company, Taiyuan, 030012, China,
E-mail: JinkaiS_un@outlook.com (corresponding author).
2
Senior Engineer, State Grid ShanXi Marketing Service Center, Taiyuan,
030012, China, E-mail: tyutxiaochun@163.com
3
Senior Engineer, State Grid ShanXi Marketing Service Center, Taiyuan,
030012, China, E-mail: 785078849@qq.com
Project Management
Received October 20025; revised December 24, 2025; accepted March 22, 2026
Available online May 29, 2026
Abstract: Anomalous sensitive power marketing data can significantly impact the interests of various stakeholders in the development of power enterprises. To improve anomaly detection in such data, we propose a method based on the Multi-Scale Convolutional Variational Autoencoder (MSCVAE). First, a multi-scale attribute matrix is constructed to represent the system’s state across varying time intervals within the multivariate time series of sensitive power marketing data. Next, a convolutional variational autoencoder functions as a generator of a reconstruction matrix from this attribute matrix, while an Attention-based Convolutional Long Short-Term Memory (ConvLSTM) is applied to find the temporal patterns. To address the class imbalance issue inherent in the sensitive power marketing dataset, a novel threshold-setting strategy derived from the confusion matrix is introduced. The proposed model is evaluated on two real-world sensitive power marketing datasets and compared with several benchmark models. The MSCVAE model exhibited superior performance relative to all benchmark models, which is evidenced by average increments of 42% in Precision, 40.9% in Recall, and 41.5% in F1-score.
Keywords: Sensitive power marketing data, multi-scale, ConvLSTM, anomaly detection.
Copyright © Journal of Engineering, Project, and Production Management (EPPM-Journal).
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
Requests for reprints and permissions at eppm.journal@gmail.com.
Citation: Sun, J., Xiao, C. and Yao, J. (2026). Multi-Scale Convolutional VAE for Anomaly Detection in Sensitive Power Marketing Data. Journal of Engineering, Project, and Production Management, 16(4), 2025-237.
