Predicting energy savings of retrofit measures for shopping centres using a Python-based multiple linear regression model
Year: 2026
Type: conference-paper
Abstract: Achieving Germany’s climate targets requires an enhancement of the residential and non-residential building stock energy-efficiency. Non-residential buildings, such as shopping centres, are characterised by a high degree of heterogeneity in building usage, envelope, and services. Standardised energy-efficient retrofitting strategies are often unsuitable for these building types. Therefore, resource-intensive dynamic building simulation models are frequently used to estimate the energy and carbon saving potentials. Predictive data-driven models for type-specific non-residential buildings represent a promising approach to reduce the modelling and simulation effort significantly while providing a robust basis for decision-making for energy-efficient retrofitting. This research gap is addressed in this paper. For this purpose, seven dynamic building models of shopping centres are developed in IDA ICE and validated with operational data. In addition, a reduced parameter set is identified as part of a statistical sensitivity analysis with key impact on expected savings. A data-driven multiple linear regression (MLR) model using the reduced parameter set is developed in Python. Despite the limited data set and the associated uncertainties, the MLR model successfully provides estimates in one test data set. In summary, the MLR model provides an initial low-cost estimate of non-residential building retrofit savings for planners and energy consultants.
Source: Proceedings of BauSIM
Institution: RWTH Aachen University
Language: English
Subfield: Building and Construction
Field: Engineering
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W7224765649