First Step in Real-Time Energy Management: Forecasting Energy Production Using Machine Learning Models

• Main Authors: Tercha, Wassila, Tadjer, Sid Ahmed, Chekired, Fathia, Canale, Laurent
• Format: Conference Proceeding
• Language: English
• Subjects: Boosting; Energy management systems; Feature extraction; Forecasting; forecasting model; Nearest neighbor methods; photovoltaic power prediction; power generation; Predictive models; Random forests; Reliability; renewable energy; Solar power generation; Time series analysis
• ISSN: 2994-9467
• DOI: 10.1109/EEEIC/ICPSEurope61470.2024.10751102

The industrial revolution has led to a rise in energy needs and the demand for better control strategies. Intelligent Energy Management Systems (IEMS) and the Internet of Energy (IoE) offer promising solutions to cut down energy usage in the construction industry. This paper looks at the first step before combining these ideas to address challenges in energy management which is the prediction of the source's power, particularly around non-linear functions and choosing energy sources. The study compares three machine learning models K-Nearest Neighbors (KNN), Random Forest, and Gradient Boosting - using a solar power generation dataset. Results show that the Random Forest model performs better than the others in predicting power generation patterns, demonstrating robust performance. However, the limitations of the Gradient Boosting and KNN models highlight the importance of selecting appropriate models tailored to specific tasks. Overall, this research provides insights to enhance energy management systems for improved efficiency and cost-effectiveness.