Digital representations of buildings play a crucial role in building retrofitting and energy modelling. Acquiring sufficient and accurate building data poses challenges, particularly in obtaining geometric information about a building's roof. Previous studies have been conducted on roof detection using aerial laser scanning point cloud data on an urban city scale. However, extracting roof geometry information from individual buildings using terrestrial laser scanning remains a challenging task due to the natural incompleteness of data for building roofs. This research aims to propose a digital twin-based framework for automatically generating the geometric information of existing building roofs required for retrofitting from terrestrial laser scanning point cloud data. The 3D laser scanning technique is adopted for data capturing. The framework includes: (1) collecting and pre-processing the point cloud data, (2) automated detection of the roof points from the building's point cloud, and (3) extracting roof geometry parameters required for building energy modelling. The final output of the proposed method is an information-rich digital twin of the building's roof. The proposed framework is validated through a case study, presenting an automated pipeline to enrich a geometric digital twin of the roof with details. This research contributes to providing geometric information for energy simulation, building energy modelling, and building retrofitting. The findings can be further validated across various building types and applied to urban building energy modelling.