Research on intelligent analysis of low voltage and multidimensional anomaly identification method driven by big data

View at publisher Summarise with VIRAT
Authors: Zhanao Wang, Lei Wang, Yue Zheng
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In the field of electric power system (EPS), it is of great significance to ensure stable power supply and improve power supply quality. As a key factor affecting power supply, low voltage problem has obvious limitations in traditional analysis methods. Therefore, this article is devoted to the research of low-voltage intelligent analysis and multi-dimensional anomaly identification method driven by big data. This article uses big data to process the massive operation data collected by EPS, and through data preprocessing, feature extraction and other steps, constructs an intelligent analysis model of low voltage based on LSTM (Long-term and Short-term Memory Network) and a multi-dimensional anomaly identification model based on SVM (Support Vector Machine). The results show that, in the intelligent analysis of low voltage, the accuracy of LSTM in predicting low voltage on the test set reaches 92%, the recall is 88%, and the RMSE (Root Mean Square Error) is 0.035, which is better than the traditional statistical analysis model. In multi-dimensional anomaly recognition, the accuracy of SVM model is 90%, and the false alarm rate is as low as 8%, which has obvious advantages compared with decision tree model. The proposed method effectively improves the accuracy of low voltage analysis and the ability of multi-dimensional anomaly identification, and provides support for the stable operation of EPS.

Fields of Study & Subjects

Smart Grid and Power Systems Power Quality and Harmonics Energy Load and Power Forecasting

Publication details

Journal / source
IET conference proceedings.
Volume
2025
Issue
47
Pages
1129–1135
Publication date
1 March 2026
Language
EN