An Automated Machine Learning and Analytics Framework for Data-Driven Optimization of Multi-Source Regional Energy Systems

Abstract

Abstract With rising energy demands and sustainability goals, optimizing a region's energy mix is critical but challenged by data complexities. Existing solutions face limitations in siloed data sources, inaccurate forecasting, static visualizations, and manual modelling workflows. This pioneering research develops ENERLIZER, the first open-source solution to integrate real-time data analytics and machine learning for end-to-end automated insights into robust multi-energy optimization. This solution is developed on the Streamlit framework and leverages integration with real-time databases to enable continuous operationalization and rapid adaptation. It provides an intuitive interface that transforms raw dataset inputs through customized pipelines. Automated descriptive analytics and aggregations are performed using Pandas. Custom XGBoost Models enable precise energy generation and demand forecasting tailored to the data. Interactive location-based visualizations are generated using Plotly. Validated using data from Spain's energy portfolio from 2015-2018, ENERLIZER delivers more than 600% reduction in analysis time versus conventional manual methods, while improving forecast accuracy. Comprehensive analytics and visualizations provide holistic insights for robust optimization. By pioneering real-time integrated data analytics, machine learning and automation, ENERLIZER breakthroughs key limitations in current energy optimization approaches. This enables more rapid, accurate and holistic data-driven decision-making for strategic energy mix planning and management.