The energy consumption prediction of heating ventilating and air-conditioning (HVAC) systems in public buildings is essential for promoting energy efficiency. However, HVAC energy consumption often fluctuates significantly due to weather variations and occupancy uncertainties within public buildings. To address this issue, this paper introduces a probabilistic prediction model based on the deep learning fusion model to quantify the energy consumption ranges with specific confidence intervals. First, the impact of temporal and environmental characteristics on HVAC energy consumption are analyzed to select relevant features. Second, we combine Long Short-Term Memory with Conformalized Quantile Regression model to obtain prediction intervals of energy consumption. Finally, to improve the model's generalization performance, an ensemble learning method is introduced to adapt varying time-series lengths through homogeneous model enhancements. Based on realistic data from an office building in the University of Macau, case studies validate the superior accuracy and generalization of the proposed model.