Skip to content
Open access

Voltage Anomaly Detection in Real-World Cloud-Based BMS Data via Context-Enhanced Multiscale LOF and Battery-Model Residual Fusion

Oct 2026 · Batteries · Vol 12, pp. 393 · 0 citations · 27 references

TL;DR

This paper proposes a method that combines a context-enhanced two-stage ensemble local outlier factor (CE-TELOF) with voltage prediction residuals generated by a composite battery model, and shows that multiscale configurations provide more stable CE-TELOF performance than single-scale configurations.

Abstract

The actual voltage data collected by the cloud-based battery management system (BMS) are susceptible to disturbances caused by sensor failures, electromagnetic interference, and communication anomalies, which may lead to four types of anomalies: random point anomalies, continuous segment anomalies, fixed-bias anomalies, and data-freezing anomalies. These anomalies reduce the reliability of battery state estimation and delay early fault warning. In order to improve the reliability of voltage anomaly detection, this paper proposes a method that combines a context-enhanced two-stage ensemble local outlier factor (CE-TELOF) with voltage prediction residuals generated by a composite battery model. CE-TELOF characterizes local abrupt changes and persistent anomaly patterns through multidimensional slope statistics, Gaussian context fusion, multiscale local outlier factor analysis, and overlapping-window coverage consistency. Meanwhile, the composite battery model provides physically interpretable evidence of anomalies by evaluating the difference between the measured voltage and the voltage predicted by the battery current–voltage dynamic model. Its parameters are identified through a two-stage optimization strategy combining a genetic algorithm and nonlinear least squares optimization. The two branches are robustly standardized and adaptively weighted to obtain the final anomaly score. The identified composite battery model achieves R2 values exceeding 0.998 across the three datasets, with consistently lower estimation errors than the conventional first-order RC model. Meanwhile, sensitivity analysis of the neighborhood-scale set shows that multiscale configurations provide more stable CE-TELOF performance than single-scale configurations. Across three real-world datasets and four anomaly types, the proposed detection method achieves F1-scores ranging from 0.8140 to 0.9836 and effectively reduces false positives associated with normal operating-condition transitions. In addition, validation using Vehicle No. 9 from a publicly available real-world on-road EV dataset and the NASA B0005 battery dataset further demonstrates the applicability of the proposed method across different data sources and battery operating conditions.

Read PDF

Similar papers

Open access Sep 2026

Advanced data driven models based on machine learning for detection of faults and failures in solar based renewable energy systems

A novel framework that integrates Extended Kalman Filter (EKF) state estimation with uncertainty-aware graph learning for photovoltaic (PV) array fault detection and localization and demonstrates strong robustness to sensor noise and transient faults by leveraging physical uncertainty to guide graph topology.

Saud Wasly, N. Abu-Hamdeh · 0 citations
Open access Oct 2026

Time-Series Anomaly Detection in Off-Grid Solar-Powered Security Systems Using Deep Learning Approaches

Reliable operation of off-grid solar-powered security systems is essential for maintaining uninterrupted surveillance and monitoring in remote locations. This study proposes a fully unsupervised deep learning framework for detecting energy-related anomalies in multivariate solar-system time-series data before they affe...

Birol Şen, Fırat Hardalaç, Kubilay Ayturan · 0 citations
Open access Oct 2026

An Edge–Cloud Collaborative Framework for Anomaly Detection and Hierarchical Storage in Concrete Bridge Sensor Networks

Distributed monitoring systems generate massive high-frequency data, posing challenges for real-time processing and storage. Existing edge–cloud methods lack a hierarchical mechanism from noise filtering to feature-space anomaly detection and may struggle to identify subthreshold deviations. This paper proposes an edge...

Su-Lan Li, He-Yao Chen, Wei Liu et al. · 0 citations
Open access Sep 2026

Kalman Filter–Enhanced Threshold Framework for Strong Wind Anomaly Detection

The detection of strong wind anomalies is crucial for meteorological monitoring, environmental assessment, and mitigating wind-related risks. However, conventional threshold-based methods often lack robustness when processing noisy and unstable wind-speed data collected under dynamic outdoor conditions. To address this...

Kartika Sari, Rahmi Hidayati, N. Huda et al. · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.