Predictive Digital Twins for Energy Consumption and Indoor Environment Quality Forecasting
737 Downloads
Last Updated: 7-2025
This research presents the development of a predictive Digital Twin system designed to forecast energy consumption an Indoor Environment Quality (IEQ) in real-time. Digital Twins can offer promising insight into a building’s energy consumption and IEQ by integrating Building Information Modelling (BIM), IoT sensor data and machine learning models. Despite this potential, many current implementations of Digital Twins do not include real-time predictive capabilities. This research aims to address that gap through the development of a predictive Digital Twin, integrating machine learning forecasting models, real-time data streams and a user interface. Literature research identifies black-box forecasting models, especially Neural Networks, as the most effective approach and highlights current challenges in data integration and visualization.
Building on the insights from the literature research, a case study is conducted on Building 59 at the Lawrence Berkeley National Laboratory, an office building located in California, USA. Historical sensor data including temperature, humidity, CO2 levels, energy loads and weather conditions are used to train and evaluate various forecasting models, including various types of Neural Networks, Residual Networks and Ensemble models. It was found that Residual Networks consistently achieve the best performance. The models also demonstrated transferability when applied data from different contexts.
A near real-time data stream was set up using the MQTT protocol. The best performing models were applied to this stream, enabling live forecasts. A dashboard was built, which displays the near real-time data and predictions through visualizations, including graphs and a color coded floorplan. The results show that predictive Digital Twins can be effectively realized, supporting monitoring of buildings and proactive decision-making. Future research should focus on improving the Digital Twin system architecture further and integrating with smart building installations to automate responses to the predictions and insights.
