Error Distribution-based Anomaly Score for Forecasting-based Anomaly Detection of PV Systems

For forecasting-based anomaly detection for PV systems, in this paper, we propose a way for calculating anomaly scores. The basic idea of our approach is to utilize the distribution of forecasting errors of normal data to derive relative anomaly score, which is limited to from 0 to 100. To further i...

Full description

Saved in:
Bibliographic Details
Published in2021 International Conference on Information and Communication Technology Convergence (ICTC) pp. 1144 - 1146
Main Authors Lee, HyunYong, Kim, Nac-Woo, Lee, Jun-Gi, Lee, Byung-Tak
Format Conference Proceeding
LanguageEnglish
Published IEEE 20.10.2021
Subjects
Anomaly detection
Anomaly score
Buildings
Convergence
error distribution
Forecasting
forecasting-based
Information and communication technology
PV systems
Online AccessGet full text

Cover

More Information
Summary:For forecasting-based anomaly detection for PV systems, in this paper, we propose a way for calculating anomaly scores. The basic idea of our approach is to utilize the distribution of forecasting errors of normal data to derive relative anomaly score, which is limited to from 0 to 100. To further improve the anomaly score, we apply our basic idea to each month separately because the distribution of forecasting errors changes over time. Through experiments using the real data, we examine some aspects of our approach preliminarily.
DOI:10.1109/ICTC52510.2021.9620808