Energy Analysis Of Community Buildings Based On Forward Uncertainty Analysis

With the rapid expansion of urban construction scale in China,energy consumption caused by urban buildings is gradually increasing,which leads to a series of environmental problems,including greenhouse effect and smog.Because building energy use has high inherent uncertainty,and the factors causing uncertainty in building energy consumption are often diverse,it is necessary to explore the impact of various factors on building energy uncertainty in order to reduce energy use and improve the quality of urban life in urban environments.Forward uncertainty analysis can consider variations of energy models in assessing urban building energy performance.The simulation study of urban building energy consumption can be used to thoroughly investigate the characteristics of energy use of urban buildings and also consider the heterogeneity between buildings.Based on the constructed geographic information system(GIS),this study would realize the automation construction of building energy consumption models in the TEDA(Tianjin Economic-Technological Development Area)campus of Tianjin University of Science and Technology.This study also proposes a set of forward uncertainty research method for urban buildings,including the construction of GIS database,automated modeling,selection of machine learning models,forward uncertainty analysis based on machine learning models,and two-dimensional Monte Carlo forward uncertainty analysis.Firstly,the data types and construction method is proposed to create an urban geographic information database.which can be used for further analysis of energy use in urban buildings.Secondly,the automated modeling of building energy consumption model is completed by R language based on the constructed geographic information database,and the automatic modeling of urban building is realized to create thousands of building energy models.Thirdly.the training set is used to construct five machine learning models,and the predictive capabilities of these machine learning models are compared by using a test set to find the optimal models.The results indicate that the MARS(multivariate adaptive regression splines)and BMARS(bagging MARS)machine learning models perform best in this study,and the Latin hypercube sampling(LHS)has better convergence than simple random sampling,and d?e sampling number of 100 in using the LHS technique is a good balance between computational cost and model accuracy.Finally,the two-dimensional Monte Carlo forward uncertainty analysis was carried out on four types of buildings in TEDA campus of Tianjin University of Science and Technology using MARS and BMARS models.The effects of both epistemic uncertainty and aleatory uncertainty on building energy consumption data were considered,respectively.The performance indicators are annual cooling energy consumption,annual heating energy consumption,and annual electrical energy consumption normalized by the floor area.The results indicate that the impact of aleatory uncertainty on building energy consumption is greater than that of epistemic uncertainty.The epistemic uncertainty has less influences on the cooling energy consumption of dormitory buildings and the aleatory uncertainty has greater impact on the heating energy consumption for classroom buildings.It is also found that the epistemic uncertainty has greater impact on the electricity energy consumption for classrooms in this research.This study forms the solid foundation for the analysis of energy consumption of-existing buildings in urban areas and also promotes the development of smart cities and building information models.Moreover,this research would also provide reliable guidance for energy conservation and emission reduction of urban buildings in China.