INTELLIGENT SYSTEM FOR ENERGY CONSUMPTION FORECASTING BASED ON MONITORING OF INDUSTRIAL EQUIPMENT USING ML ALGORITHMS

Abstract

The article presents a software solution that enables real-time energy consumption forecasting and monitoring of the technical condition of industrial equipment. Unlike traditional systems that rely on averaged statistical data, the proposed approach integrates machine learning methods with telemetry analysis, allowing for more accurate predictions, timely detection of equipment deviations, and prompt response to anomalies. The purpose of the study is to develop an intelligent integrated solution that combines analysis of the technical state of industrial equipment with energy consumption forecasting based on collected telemetry. This approach ensures improved energy efficiency and reduces the financial losses of enterprises. Methodology. The study relies on time series of energy consumption and telemetry parameters of industrial equipment (temperature, vibration, workload). The data undergo preprocessing (noise filtering, normalization, aggregation) and are then processed by the LightGBM gradient boosting algorithm, which ensures accurate real-time forecasting. The system architecture is implemented using a microservice approach with Docker containers, MongoDB for data storage, and a Flask-based dashboard for visualization. Such an architecture allows for flexible scaling and adaptation of the system to various industrial conditions. Scientific novelty. The research introduces a multi-component intelligent system that simultaneously performs monitoring, forecasting, and energy management functions based on machine learning algorithms and modern telemetry data processing technologies. Unlike existing solutions, the proposed system integrates predictive models with real-time equipment condition monitoring, enabling anomaly detection and automated analytical reporting. Conclusions. The developed system demonstrates the potential to reduce electricity costs, increase the level of automation in resource management, and improve the efficiency of industrial processes. Future work involves implementing self-learning mechanisms, expanding the set of telemetry parameters, and achieving full integration with energy market platforms to enable automated surplus energy trading.