Building energy demand modeling: from individual buildings to urban scale

Abstract

This chapter explores the complexities and methodologies of building energy demand modelling, tracing the transition from individual structures to the urban scale. Given that the building sector is responsible for approximately 35% of the world’s final energy consumption, the text emphasises the necessity of Building Energy Models (BEMs) for predicting performance indicators and developing sustainable policies. The authors focus primarily on bottom-up approaches, which utilise disaggregated data to model multiple entities before extrapolating results to an entire district or city. Within this framework, three main modelling categories are identified: physics-based (White Box), data-driven (Black Box), and hybrid (Grey Box) approaches. The chapter highlights critical data challenges at the urban scale, including issues with data availability, uncertainty, and the significant computational time required to simulate vast numbers of buildings. To address these hurdles, the chapter details various simplification and optimisation techniques: 1. Clustering Methods: Unsupervised learning algorithms are investigated as a means to identify representative buildings (archetypes), effectively reducing the modelling burden by grouping similar structures based on geometric and energy-related features. 2. Reduced-Order RC Models: These models employ a thermoelectric analogy, using thermal resistances (R) and capacitances (C) to simplify heat transfer equations, offering a flexible balance between accuracy and computing time. 3. Electrical Appliance Modelling: A novel multistate survival analysis approach is introduced to model the stochastic usage of low-load appliances, allowing for the prediction of realistic energy demand profiles for communities. 4. District Demand Profiling: The text describes DiDeProM, a model that generates detailed thermal energy demand profiles using parametric analysis on reference buildings, alongside data-driven emulators like neural networks that can provide rapid demand estimations. Ultimately, the chapter asserts that integrating these advanced modelling strategies provides essential decision support tools for urban planners and stakeholders aiming to mitigate environmental impacts and design low-carbon cities.