A study on hybrid machine learning models with data decomposition and similar days selection method for daily peak load forecasting: application to Thailand’s electricity demand

Accurate electricity demand forecasting is pivotal for the efficient and sustainable operation of modern power systems. This research addresses critical challenges in load forecasting by developing a novel hybrid forecasting model that integrates advanced decomposition techniques (Variational Mode Decomposition- Empirical Mode Decomposition-Fast Fourier Transform), input variable selection methods (Stepwise Regression and Similar Day Selection), and Artificial Neural Network. The proposed model is designed to overcome the inherent nonlinearity and non-stationarity of electricity load data while optimizing computational efficiency. Key contributions of this study include the refinement of input variable selection, enabling the model to focus on the most influential factors, and the integration of decomposition techniques to enhance scalability and robustness. Moreover, the proposed model is specifically designed to address the imbalance between normal days and special holidays, enhancing the efficiency of the forecasting model. This is particularly important for special holidays, which exhibit distinct patterns compared to regular weekdays and weekends. Using Thailand's electricity demand as a case study, the research validates the model's practical applicability and demonstrates its superiority over existing forecasting models, particularly in handling challenging periods such as special holidays. Experimental results show that the proposed model delivers superior performance while significantly reducing computation time by addressing critical issues in input variable selection, data decomposition, and imbalanced data between normal and special days during the training process. This work is expected to improve grid stability and enable more accurate load scheduling, with significant implications for policymakers and grid operators. By advancing theoretical and practical knowledge, the study contributes to addressing global energy challenges and fostering sustainable development.