Energy Management Models in Heavy Industries: A Data-Driven Approach for Sustainability in the Chemical, Oil, and Gas Sectors

Sammanfattning

This thesis explores data driven energy management models to support industry toward a future of sustainability in heavy energy consuming and environmentally impact industries: chemical, oil, and gas. The study uses machine learning techniques to create a predictive model that can serve to maximize energy usage and minimize greenhouse gas emissions. A real world case study of a refinery in Thessalonika is then given with analyses of energy consumption patterns utilizing real world data from Hellenic Energy Sustainability Reports by analyzing operational schedules, efficiency of equipment and fuel type. Operational patterns is found to contribute up to 35% to energy use variability, which is found by way of statistical methods including correlation analysis and feature importance via Random Forest. Through strategies like predictive maintenance and load shifting, the proposed models have achieved potential energy saving in the range of 12%-18%, which falls in line with the established global sustainability goals, including targets laid in the Paris Agreement. The results represent estimated optimisation potential under modelled scenarios and are subject to limitations related to data availability, single-case study design, and the use of observational data. Suggestions for integrating artificial intelligence into an industrial energy system are offered based on the findings, as operation becomes more efficient and environmentally accountable. Based on this research, advances in energy optimization theory are made, and energy intensive sectors are provided with scaling solutions to their energy usage, a necessary step toward a future industrial sustainability