Energy forecasting models using different algorithm and modeling of hybrid renewable resources (HRE) for educational building: a comprehensive study

• Published in: Electrical engineering Vol. 107; no. 5; pp. 6305 - 6328
• Main Authors: Thulasingam, Muthukumaran, Periyanayagam, Ajay D. Vimal Raj
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.05.2025; Springer Nature B.V
• Subjects: Algorithms; Alternative energy sources; Business metrics; Data points; Deep learning; Economics and Management; Education; Electrical Engineering; Electrical Machines and Networks; Energy consumption; Energy demand; Energy Policy; Energy resources; Engineering; Ensemble learning; Errors; Feasibility studies; Forecasting; Forecasting techniques; Humidity; HVAC; Machine learning; Neural networks; Original Paper; Photovoltaic cells; Power Electronics; Python; Radiation; Real time; Regression models; Relative humidity; Renewable energy sources; Renewable resources; Research methodology; Statistical analysis; Support vector machines; System reliability; Time series; Wind speed; Energy demand; HRE; R metrics; HOMER Pro; Energy forecasting
• ISSN: 0948-7921; 1432-0487
• DOI: 10.1007/s00202-024-02847-1

Forecasting energy has become crucial in modern power systems to ensure efficient operation. Enhanced forecasting tools enable accurate prediction of load and energy demand well in advance, ensuring system reliability. This study focuses on predicting the energy demands of educational institutions, employing various models such as neural networks, support vector machines, ensemble methods, and machine learning algorithms. Using real-time energy data from an educational institute in Hyderabad, consisting of nearly 8760 data points (365*24), forecasting models were trained. The performance of each model was evaluated using R metrics including mean absolute error (MAE), mean absolute percentage error (MAPE), mean squared error (MSE), root-mean-squared error, and R-squared ( R 2 ). Among the eight models developed, the random forest regression model exhibited superior accuracy with the lowest R metrics values. For the random forest regression model, the metrics were as follows: MSE 0.053, MAE 0.117, MAPE 0.0281, and R 2 score 0.9995. This model was trained with four input features (hours, temperature, wind speed, and relative humidity) and one output (energy). Utilizing this model, energy predictions can be made for any day, hour, month, or year. Furthermore, the predicted energy data from the model were employed in the modeling of hybrid renewable energy sources (HRE) systems to satisfy the building’s power demands. A techno-economic feasibility analysis of the hybrid renewable energy (HRE) system was conducted, optimizing the PV and wind sources to 133 kW and 1 kW, respectively. This resulted in a minimum cost of energy (COE) of 0.009 USD and an 81% reduction in CO 2 emissions.