کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

Machine Learning-Based Classification of Electricity Load Levels for Short-Term Energy Management

نویسندگان :

Javad Hajiannezhad¹ Habib Rajabi Mashhadi² Maryam Baradaran Naseri³ Mohammadreza Mohammadhasani⁴

1- دانشگاه فردوسی مشهد 2- دانشگاه فردوسی مشهد 3- دانشگاه صنعتی شریف 4- دانشگاه فردوسی مشهد

کلمات کلیدی :

Electricity demand classification،load level prediction،machine learning،smart meter data،residential energy analysis،short-term forecasting

چکیده :

With the increasing integration of smart grids and smart meters, vast amounts of consumption data are now available for analysis. Instead of forecasting exact numerical values, classifying electricity demand into discrete load levels can support practical decision-making in demand response, energy management, and tariff optimization. In this paper, a machine learning-based approach presented for classifying electricity consumption into three levels: low, medium, and high. Real residential smart meter data from multiple locations is used, along with time-based and weather-related features. Load levels are assigned using quartile-based thresholds. A variety of classification algorithms, including logistic regression, decision trees, ensemble models, and generalized additive models examined. Random Forest and Bagging classifiers achieve the best performance, with over 87% test accuracy. Confusion matrix analysis and feature importance scores are also provided to support interpretability and robustness. The results demonstrate that machine learning methods can effectively categorize short-term electricity usage, offering a reliable tool for energy-aware planning in smart grid environments.