Towards efficient energy systems: Improving hourly load forecasting for smarter management in Albania

Effective energy load prediction enhances the integration of renewable energy sources that supports the demand response and lead to more sustainable and cost-effective energy systems. This study contributes to the development of a smart energy system in forecasting hourly electrical load in Albania. The data taken into consideration fall in a timeline of three years. The main goal is to improve energy efficiency and optimize resource management by developing accurate and data-driven forecasting models. Accurate and reliable load forecasts are essential to optimize energy distribution, reduce operating costs and maintain grid stability, all of which are essential to building a sustainable and efficient energy infrastructure. Traditional time series models, such as ARIMA and SARIMA, have been widely utilized in load forecasting due to their straightforward design and robustness; however, advancements in machine learning now provide new methods capable of capturing complex patterns and long-term dependencies within time series data. This study applies both classical time series and machine learning models, especially the LSTM model, incorporating exogenous variables such as temperature to improve the precision of short-term and medium-term predictions. Model performance is evaluated using accuracy metrics, revealing that machine learning models generally outperform classical time series models, and produce more accurate and reliable results. The findings of the study provide practical insights for energy planners, grid operators, and policymakers in country and region, offering them advanced, precise tools to optimize energy use, enhance grid management, and support efficient resource allocation. The study's key contribution lies in demonstrating the effectiveness of data-driven methods for Albania’s energy sector, presenting an application that supports a more resilient and sustainable energy ecosystem.