An Operational Carbon Emission Prediction Model Based on Machine Learning Methods for Urban Residential Buildings in Guangzhou

Abstract

The carbon emissions of urban residential buildings are substantial. However, the standard operating conditions specified in current energy-saving standards significantly differ from the actual energy consumption under real operating conditions. Therefore, it is essential to consider the impact of residents’ actual energy consumption behavior in carbon emission forecasts. To improve the accuracy of carbon emission predictions for urban residential buildings, this paper focuses on residential buildings in Guangzhou. Taking into account the energy consumption behavior of residents, parameterized modeling is carried out in the R language, and simulation is carried out using EnergyPlus software. Analysis revealed that the higher the comfort level of residential energy consumption behavior, the more it is necessary to encourage residents to adopt energy-saving behaviors. Combining carbon emission factors, air-conditioning energy efficiency, and the power consumption models of lighting and electrical equipment, a comprehensive operational carbon emission prediction model for urban residential operations in Guangzhou was developed. By comparing the prediction model with an actual case, it was found that the prediction deviation was only 4%, indicating high accuracy. The proposed operational carbon emission model can quickly assist designers in evaluating the carbon emissions of urban residential buildings in the early stages of design, providing an accurate basis for decision-making.

Keywords

• operational carbon emission prediction
• urban residential buildings
• sensitivity analysis
• energy consumption behavior
• machine learning method