Prediction of Indoor Air Temperature Using Weather Data and Simple Building Descriptors

Area of Science:

• Environmental Health
• Building Science
• Climate Change Adaptation

Background:

• Non-optimal air temperatures pose significant risks to human health and productivity.
• Climate change is increasing the frequency and intensity of extreme heat and cold events.
• Existing methods for assessing indoor thermal environments require improvement for personalized interventions.

Purpose of the Study:

• To develop a predictive model for indoor air temperature to assess thermal indoor environments.
• To create a method for a smartphone application (ClimApp) providing individualized advice on thermal stress.
• To evaluate the accuracy of a predictive model using weather data and building attributes.

Main Methods:

• A decision tree classification algorithm was employed to predict discrete indoor air temperatures.
• Input data included online weather services and general building attributes provided by users.
• The model was trained and tested using field measurements from seven Danish households and building simulations across three climate regions.

Main Results:

• The predictive method achieved 92% accuracy (F1-score) when predicting temperatures under known conditions (same household, occupants, climate).
• Performance decreased to 30% accuracy when applied to different climate conditions.
• The model demonstrated higher accuracy in predicting commonly observed indoor temperatures.

Conclusions:

• A straightforward and reasonably accurate method for indoor temperature estimation is feasible using weather data and basic building attributes.
• The developed method shows potential for integration into smartphone applications to provide personalized thermal comfort advice.
• Further research is needed to improve model performance across diverse climatic conditions for broader applicability.