Electric Load Forecasting Using Time Series Models: Comparative Studies

Abstract Accurate electric load forecasting plays a crucial role in ensuring the smooth operation and strategic planning of power systems. Reliable forecasts help utilities make informed decisions about generation scheduling, grid stability, and resource allocation. Traditional forecasting methods, such as statistical time series models, often depend heavily on large amounts of historical data. While effective in stable conditions, their accuracy tends to decline for long-term predictions, especially when recent data is limited or no longer reflects current patterns. In this paper, we present a detailed comparative analysis of a range of forecasting techniques, including regression-based methods, classical machine learning algorithms, and advanced deep learning architectures such as Long Short-Term Memory (LSTM) networks and Transformer models. These models are evaluated using real-world electric load data from the Sultanate of Oman covering the years 2020 to 2022 for training and validation, with 2023 data held out for independent testing to ensure an unbiased assessment. Our findings show that deep learning models particularly LSTM networks consistently deliver outstanding accuracy and robustness, achieving testing R 2 scores above 0.98 and mean absolute percentage errors (MAPE) as low as 2.58%. Boosted ensemble methods also perform exceptionally well, reaching a testing MAPE below 1%, which makes them highly practical for operational use. In contrast, traditional time series models such as Autoregressive Integrated Moving Average (ARIMA) and its seasonal variant (SARIMA) generally produce lower accuracy and higher error rates, especially in testing scenarios. This study highlights the strengths and limitations of various forecasting methods, guiding practitioners in selecting models that best fit their data and needs. The findings show how modern ML and DL techniques can boost accuracy and support more reliable, efficient load management for the Omani power grid.