Accurate power demand prediction is essential for maintaining supply-demand balance, reducing energy costs, and supporting proactive building energy management. Machine learning approaches have demonstrated strong performance in predicting energy and power demand. However, they typically rely on abundant high-quality historical data, which are often unavailable for newly monitored buildings. To address the challenge of low prediction accuracy under data scarcity, this study explores the potential of hierarchy learning (HL) to improve overall and peak power demand prediction with limited data. First, a similarity analysis method, Dynamic Maximum (DM), combining dynamic time warping and maximum mean discrepancy, is developed to identify the source domain data. The selected source and target domain data are then used within an HL framework to train a decomposition-based prediction model, namely a multiple seasonal-trend decomposition using LOESS combined with a Bayesian additive regression tree (MSTL-BART), forming the proposed DM-MSTL-BART-HL framework. The framework is evaluated using real-world data from residential buildings. Our results demonstrated that the framework significantly reduced CV-RMSE error with only one week of target domain data, achieving predictive accuracy improvements ranging from 2.58% to 19.18% compared to the baseline model, ASHRAE requirements, and other state-of-the-art methods. For peak demand prediction, the framework achieved CV-RMSE values ranging from 2.96% to 21.08%, with corresponding coverage width criterion values between 0.24 and 1.23, demonstrating a strong balance between prediction interval width and coverage. This method highlighted the potential of DM-MSTL-BART-HL as a practical solution for building power demand prediction when sufficient historical data are not available.

Accurate power demand prediction is essential for utility operations and building energy management to enable efficient grid operation and proactive control of energy storage systems. In residential buildings, demand patterns are highly variable due to diverse occupant behaviors and usage fluctuations, introducing uncertainty that makes accurate forecasting challenging. However, existing approaches often focus on point predictions that lack demand uncertainty information, limiting the ability to characterize uncertainty to make appropriate operational decisions. One potential approach is to employ an accurate and reliable probabilistic model. In this context, this study develops a probabilistic forecasting framework to capture uncertainties in the applications of predicting power demand at both single- and multi-unit levels. First, a multiple seasonal-trend decomposition using the LOESS algorithm decomposes power consumption data into trend, seasonal, and residual components to mitigate the effects of noise. This procedure enables the predictive model to focus on the meaningful variations that affect demand. Then, we extend the Bayesian additive regression tree technique by incorporating both linear and non-linear components to capture complex relationships and generate probabilistic forecasting. The proposed method is evaluated using real-world datasets from residential buildings, focusing on both overall and peak demand scenarios. The power demand data exhibit high variability and complexity, representing diverse occupant behaviors and usage patterns. Results show that the proposed method achieves high accuracy for both mean and probabilistic predictions. It achieves an average CV-RMSE of 15.48% for overall predictions and 16.12% for peak demands, along with PICP values of 97.91% and 97.94%, respectively. These results significantly exceed ASHRAE requirements and outperform other benchmark methods. The findings show that the proposed method can support risk-informed decision-making and enhance energy efficiency.