Skip to content
Conference

Big-data mining-based identification of abnormal electricity consumption behavior for urban transport-energy distribution networks

Sep 2026 · International Conference on Intelligent Transportation Systems and Automation Control · Vol 14368, pp. 143681Z - 143681Z-10 · 0 citations · 17 references
Engineering

TL;DR

A big-data mining framework that combines chronological AMI cleaning, feeder-relative consistency analysis, temporal convolutional representation, gradient-boosted tabular learning and isolation-based anomaly scoring is proposed that provides a reproducible decision-support approach for abnormal-consumption identification in urban low-voltage networks.

Abstract

Urban distribution networks that support residential areas, commercial districts and electric vehicle (EV) charging facilities generate large volumes of advanced metering infrastructure (AMI) data. Abnormal electricity consumption behavior, including meter bypassing, partial load scaling, flat-line reporting and false data injection, causes non-technical loss and can also disturb automated feeder operation. This paper proposes a big-data mining framework that combines chronological AMI cleaning, feeder-relative consistency analysis, temporal convolutional representation, gradient-boosted tabular learning and isolation-based anomaly scoring. The benchmark uses the UK Power Networks Low Carbon London dataset, which contains half-hourly electricity-use records from 5,567 Greater London households collected between November 2011 and February 2014. A fully specified 24-month interval from 1 March 2012 to 28 February 2014 is aggregated to hourly resolution and divided chronologically into 438 training days, 146 validation days and 146 test days. Six abnormal consumption templates are injected only after the split. Under this 60%-20%-20% protocol, the proposed model reaches an area under the receiver operating characteristic curve (AUC) of 0.957 and an F1-score of 0.873, outperforming logistic regression, random forest, extreme gradient boosting (XGBoost) and long short-term memory autoencoder (LSTM-AE) baselines. In a fixed 42-day holdout containing 5,567-meter identifiers and 5,611,536 expected hourly meter-time positions before quality filtering, the top-5% ranked list recovers 103 of 126 injected abnormal windows (81.7%). The framework therefore provides a reproducible decision-support approach for abnormal-consumption identification in urban low-voltage networks.

View source

Similar papers

Open access Sep 2026

Decentralized Smart Grid Load Scheduling for Commercial EV Charging Stations in Pakistan Using Machine Learning

A decentralized, machine learning-driven load scheduling architecture for Pakistani commercial EV charging stations is proposed in this study, ideal for Pakistan's limited SCADA and ICT environment because it does not require real-time cloud access.

Syed Aqeel Shah, Amjad Khattak · 0 citations
Open access Sep 2026

An Improved Informer-Based Load Forecasting Method Incorporating Electricity Consumption Profiling

An electricity-use-aware load forecasting method that integrates user behavior profiling with an improved Informer model and indicates that combining electricity-use behavior profiling with multiscale temporal modeling can effectively improve residential demand-side load forecasting.

Jin Wang, Ying Shi, Lei Zhang · 0 citations
Open access Sep 2026

Machine Learning-Based Electricity Theft Detection in Smart Grids using Class Imbalance Handling and Shapley Additive Explanations (Shap) Explainability

The results make clear that ignoring class imbalance renders theft detection systems ineffective in practice, and that SHAP explanations are a meaningful bridge between statistical performance and the operational confidence utilities need to act on model outputs.

Sena Gemechis File · 0 citations
Conference Open access 2026

Electric Vehicle Charging Demand Prediction in Shenzhen Based on Random Forest and LSTM

Reliably forecasting how much energy electric vehicles (EVs) will require at public chargers is a prerequisite for infrastructure planning and matching available capacity to real-world usage. This study compares two representative approaches, Random Forest and Long Short-Term Memory (LSTM), to forecast hourly charging...

Hao-Ran Luo · 0 citations
Open access Sep 2026

INTELLIGENT CLASSIFICATION OF RURAL AREA ELECTRICITY CONSUMPTION MANAGEMENT PROFILES USING FUZZY CLUSTERING AND MACHINE LEARNING

The rising variability in household electricity usage behaviour needs robust data-driven consumer categorisation systems. This paper suggests using actual monthly electricity consumption data to develop a Fuzzy C-Means (FCM)-based framework for categorising residential electricity consumers. The dataset includes 48 mon...

Durgadevi Velusamy, I. S, R. Kumar B 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.