DEVELOPMENT AND VALIDATION OF AN INTELLIGENT ENERGY MANAGEMENT SYSTEM FOR INDUSTRIAL ENTERPRISES USING MACHINE LEARNING

Abstract

This paper presents the development of an automated energy consumption management system (AECMS) for an industrial enterprise based on machine learning methods. A comparative analysis of load forecasting algorithms — ARIMA, SVR, XGBoost, LSTM, and Transformer — was conducted. A three-tier system architecture (Edge Layer, Analytics Layer, Decision Layer) was proposed using Apache Kafka, InfluxDB, and PostgreSQL. An energy consumption anomaly detection module based on the Isolation Forest algorithm was developed. System validation was carried out at Temirtau-Metal LLP over a 6-month period, achieving a 17.3% reduction in average monthly energy consumption and an 18.8% decrease in peak load. The total annual economic effect amounted to 12.5 million KZT with a payback period of less than 8 months