
MC2025 Abstracts & Papers
Session 3: Assessing occupants wellbeing and comfort
LSTM Forecasting of UCL Central House’s Office Temperature Using Real-Time BMS Data under Future Climate Conditions
Nakanya Nonthiworawong
This study aims to forecast future indoor thermal performance and evaluate the operational efficiency of the Building Management System (BMS) in UCL Central House (Bartlett Office). The research has two objectives: (1) to utilise real sensor data from the BMS to assess current performance and its practical use, and (2) to apply a multivariate Long Short-Term Memory (LSTM) algorithm to forecast indoor thermal performance in 2030. Although the building was retrofitted in 2010, sensor data from 2023–2025 shows free-running indoor temperatures reaching up to 31°C—exceeding the 20–25°C range targeted during the retrofit, which remained consistent in 2017—suggesting overheating risk and retrofit performance gap. Frequent indoor 27°C peaks in summer 2025 also indicate the current setpoint policy may become ineffective. Due to incomplete architectural data, this study avoids physical energy modelling and instead uses LSTM trained on outdoor temperature, seasonality, and occupancy behaviour. As many UCL offices are in older buildings with poor thermal design, data-driven forecasting is essential. Predicting future thermal conditions enables proactive retrofit planning, helping reduce unexpected energy use and carbon emissions—supporting London’s Net Zero 2030 goal.
Keywords: Building Management System, Indoor Thermal Performance Forecasting, LSTM Deep Learning, Overheating Risk, Retrofit Performance Gap Assessment.
– Theme: Assessing occupants wellbeing and comfort (post occupancy evaluation, indoor air quality, thermal, visual, acoustic, multimodal, mixed method) –

