Short- and Medium-Term Electricity Load Forecasting Using Machine Learning Models: A Case Study for Tirana
Session
Energy Efficiency Engineering
Description
Accurate electricity load forecasting is crucial for balancing production, distribution, and consumption of energy. Short- and medium-term predictions help manage demand over the coming days, weeks, and months, supporting efficient power system operation, market planning, and integration of renewable energy sources. Advanced forecasting techniques have been widely applied to large interconnected systems, often with the integration of exogenous variables in load management. This paper focuses on daily to monthly load forecasting for the city of Tirana, using real consumption and meteorological data, including temperature, heating degree days (HDD), and cooling degree days (CDD), to provide accurate predictions and insights for local energy management. A comparative study is conducted over a diverse set of models, considering classical regression methods, ensemble approaches such as decision trees, and machine learning architectures including artificial and recurrent neural networks. The models are evaluated for both short-term horizons (next day) and medium-term horizons (weekly and monthly). The data incorporate relevant meteorological variables and historical load patterns to capture the dynamics of electricity demand. Model performance shows accurate results across all tested scenarios, with differences aligned with their theoretical approach and the variables included. The findings highlight the challenges of applying different modeling techniques in small-scale electricity systems with strong weather dependencies. This study provides insights for improving operational forecasting in Albania and emphasizes the need for hybrid approaches and advanced deep learning architectures for robust medium-term performance.
Keywords:
Electricity load forecasting; Machine learning; Meteorological data; Energy management; Neural networks
Proceedings Editor
Edmond Hajrizi
ISBN
978-9951-982-41-2
Location
UBT Lipjan, Kosovo
Start Date
25-10-2025 9:00 AM
End Date
26-10-2025 6:00 PM
DOI
10.33107/ubt-ic.2025.154
Recommended Citation
Qosaj, Agresa, "Short- and Medium-Term Electricity Load Forecasting Using Machine Learning Models: A Case Study for Tirana" (2025). UBT International Conference. 8.
https://knowledgecenter.ubt-uni.net/conference/2025UBTIC/EEE/8
Short- and Medium-Term Electricity Load Forecasting Using Machine Learning Models: A Case Study for Tirana
UBT Lipjan, Kosovo
Accurate electricity load forecasting is crucial for balancing production, distribution, and consumption of energy. Short- and medium-term predictions help manage demand over the coming days, weeks, and months, supporting efficient power system operation, market planning, and integration of renewable energy sources. Advanced forecasting techniques have been widely applied to large interconnected systems, often with the integration of exogenous variables in load management. This paper focuses on daily to monthly load forecasting for the city of Tirana, using real consumption and meteorological data, including temperature, heating degree days (HDD), and cooling degree days (CDD), to provide accurate predictions and insights for local energy management. A comparative study is conducted over a diverse set of models, considering classical regression methods, ensemble approaches such as decision trees, and machine learning architectures including artificial and recurrent neural networks. The models are evaluated for both short-term horizons (next day) and medium-term horizons (weekly and monthly). The data incorporate relevant meteorological variables and historical load patterns to capture the dynamics of electricity demand. Model performance shows accurate results across all tested scenarios, with differences aligned with their theoretical approach and the variables included. The findings highlight the challenges of applying different modeling techniques in small-scale electricity systems with strong weather dependencies. This study provides insights for improving operational forecasting in Albania and emphasizes the need for hybrid approaches and advanced deep learning architectures for robust medium-term performance.