Indoor temperature forecasting systems for heat warning: A review of methods and applications

Abstract

With increasing frequency of hot days and heat waves globally, the risk of indoor overheating in buildings is growing. Overheating affects occupants’ health, wellbeing, and comfort and can lead to increased cooling-related energy consumption, and potential associated energy-related carbon emissions. Forecasts of building overheating can connect to active and passive cooling systems via smart devices, thereby offering potential to mitigate the impacts of overheating before they have a chance to occur, or warn residents of high future temperatures via Heat Health Warning Systems (HHWS). There are a range of different approaches and applications for indoor temperature forecasts, and the practical performance, implementation ability, and intended application vary. Following a structured review framework, this study synthesises the existing research on indoor temperature forecast models trained on historical data for the non-heating season with the objective of progressing towards HHWS for diverse housing and housing stocks. Results categorise main gaps between current HHWS and the advanced forecasting applications found in existing literature. While deep neural networks, hybrid models, and new grey-box models show significant potential in forecasting, their adoption in indoor HHWS is limited. Research directions identified in this review paper pave the way for an adaptable indoor HHWS and can guide future developments to enhance occupant comfort and health while improving energy efficiency in buildings.