No abstract

No abstract

High-quality data on the investigated area is crucial for modelling urban building energy demands, but its availability is often insufficient. We present an approach to acquire (i) building geometries, (ii) their ages, and (iii) their retrofit states. It consists of creating a 3D model from aerial imagery, determining building ages through machine learning, generating a simulation model based on open-source tools, and assessing retrofit states by comparing simulated temperatures with infrared thermography (IRT) measurements.The demonstration on a case study quarter in Berlin shows that heat demand results are comparable to other tools. Using machine learning is already wellsuited to close knowledge gaps regarding building ages. However, retrofit state assessment using IRT was unsatisfactory due to insufficient measurement accuracy and is envisaged for improvement in future research, along with a validation of the approach. Key Innovations• Combined workflow using remote sensing, machine learning, and data enrichment tools to collect building stock data • Application of an additional, Modelica-based simulation method to the case study quarter used by Dochev et al. ( 2020) who compare two urban building energy modelling approaches Practical ImplicationsHigh-quality urban building energy models may be generated even in case of limited input data availability prior to the study, but some open issues exists.