Multi-Step Ahead Short-Term Residential DC Load Forecasting: A Comparative Study of Ngboost-Based Algorithms

Accurate short-term residential load forecasting is critical for effective energy planning and management. In the near future, households will have separate AC & DC buses for individual loads for energy efficiency purposes. Therefore, DC load forecasting will be vital for energy management and efficiency. This study presents a comparative analysis of multiple NGBoost-based algorithms for multi-step ahead forecasting, leveraging feature selection centered on trends and periodic patterns in historical data. By combining probabilistic forecasting with state-of-the-art gradient boosting methods, the NGBoost framework adeptly captures nonlinear dependencies and complex load characteristics. In this paper, four different NGBoost algorithms have been proposed and rigorously evaluated using performance metrics such as root mean square error (RMSE), mean square error (MSE), and mean absolute error (MAE), across various forecasting horizons. The results of the comparative simulation demonstrate that these NGBoost-based approaches consistently achieve high predictive accuracy, delivering notable improvements in the prediction of DC residential loads. The findings underscore the potential of NGBoost in advancing multistep forecasting methodologies for enhanced energy management and system reliability.