Estimating the influence of building and urban form on the thermal loads of urban dwellings in the Mediterranean climate using machine learning

ABSTRACT

The paper presents a novel method of generating metamodels that can quantify the influence of urban form on the heating and cooling loads of dwellings for the Mediterranean climate. The goal of the metamodels is to inform the early urban design process with quick and accurate feedback on the energy demand of a given urban configuration. The proposed method generates training datasets through Monte Carlo dynamic energy simulations of a single thermal zone. The study focuses on the geometrical parameters that affect building energy balance at the urban scale, such as dwelling shape, compactness, adjacencies and shading, utilizing a set of morphological indicators to relate these parameters to heating and cooling loads. The study focuses on the Mediterranean climate of Thessaloniki, Greece and utilizes several open-source tools (Ladybug Tools, EnergyPlus and Scikit-learn) under a streamlined workflow. The metamodels, which are trained using the Random Forest algorithm are then validated on 12 real-world urban sites of Thessaloniki with a reported Mean Average Percentage Error (MAPE) of 12% and 15% for heating and cooling loads, respectively.

KEYWORDS:

• Heating loads
• cooling loads
• urban form
• residential
• energy plus
• machine learning
• metamodeling

Declaration of interest statement

The authors declare that there is no conflict of interest.

Additional information

Funding

This research is co-financed by Greece and the European Union (European Social Fund-ESF) through the Operational Programme “Human Resources Development, Education and Lifelong Learning 2014-2020” in the context of the project “Estimation of heating and cooling loads of residential urban blocks by using urban morphological indicators” (MIS 5047833).