Currently, the development of building energy simulation (BES) technology has extended load forecasting scale from individual buildings to the district level. However, detailed modelling methods that rely on extensive building design information and traditional load metric methods pose challenges in district-scale calculations. The urban building energy model (UBEM) has demonstrated versatility in district-level energy simulations, yet applying occupancy modelling at the urban scale remains challenging. Therefore, a combined simulation method that integrates UBEM and an building level occupancy sub-model with a simplified algorithm based on the shoebox model is proposed in this study, a library of building prototype templates obtained through a data crawling technique, and a modified Markov chain (MC) as occupancy model input. The method is applied to predict the district heating load of a district comprising 1068 buildings in Lhasa, Tibet. Results indicate a 46.4 % improvement in prediction accuracy compared to the traditional index method, providing valuable insights for district energy planning. In addition, a comparison of load results under different occupancy modes reveals the impact of occupant occupancy status on peak loads (−9.3 % to 18.9 %) and valley loads (−3.6 % to 40.7 %). Meanwhile, the difference in loads imposed by occupancy across different functional buildings is time dependent (4 % to 45 %), influenced by the complexity of occupant type, building type, and occupancy density.