Energy and Carbon Performance Analysis in a Mixed Commercial-office Building in Shenzhen Using EnergyPlus and LSTM
Jie Li *
College of Mechanical and Electrical Engineering, Shaanxi University of Science and Technology, Xi’an - 710021, China.
Lingbo Kong
College of Mechanical and Electrical Engineering, Shaanxi University of Science and Technology, Xi’an - 710021, China.
*Author to whom correspondence should be addressed.
Abstract
Energy conservation and carbon emission reduction in building have become key areas of research and practical application in China with the advancement of carbon neutrality. An energy consumption model was developed in this study based on a typical mixed-use commercial-office building in Shenzhen. EnergyPlus and Long Short Term Memory (LSTM) neural network were introduced to optimize the simulated energy consumption. Model validation yielded high accuracy, with R² exceeding 0.96 for both total building energy consumption and HVAC energy consumption, and RMSE below 0.25. The characteristic bimodal fluctuation in annual energy use was observed with the maximum monthly variation reaching 741 MWh. Ambient temperature and customer traffic were identified as the primary external driving factors, both exhibiting a significant positive correlation with energy consumption. Based on the developed building energy model, a comprehensive carbon flow framework was constructed, which clearly illustrating the emission pathways and distribution across various internal systems and functional zones. Based on the constructed energy consumption model, a photovoltaic system was simulated, resulting in an annual energy reduction of 145 MWh and a carbon emissions reduction of 1487.1 tCO₂.
Keywords: Building energy consumption, carbon emission, EnergyPlus simulation, LSTM neural network