Building energy efficiency assessment base on predict-center criterion under diversified conditions
Xuechao Liao,Yong Zhang,Xiujuan Zheng,Junlong Kang,Haoyi Zhao,Ning Wang
DOI: https://doi.org/10.1016/j.enbuild.2024.114118
IF: 7.201
2024-03-30
Energy and Buildings
Abstract:With the rapid expanding of urban economy and population, how to reduce the energy consumption and improve energy efficiency of public buildings has become a key issue for energy conservation and urban development. However, due to the diversity of building usage condition, energy consumption data often has problems with nonlinearity, outliers, and diversity, which greatly reduce the accuracy of energy consumption prediction and energy conservation assessment. To overcome the above drawbacks, a regression prediction model of building energy consumption is designed, which cluster the same type of data characteristics to form a subset of equipment through Gaussian mixture clustering, and design the specific normalized adaptive loss function to determine the optimal control parameters of regression model. Furthermore, a novel predict-center energy efficiency evaluation criterion is proposed, and the cumulative distribution pattern is further statistically analyzed. Finally, the energy benchmark evaluation framework is proposed to determine the energy efficiency evaluation grade. Through comparative experimental analysis, it can be seen that the proposed prediction model can adapt to different types of building data, and have better prediction accuracy and convergence speed. The consistency and robustness of the benchmark evaluation framework and method are optimal.
energy & fuels,construction & building technology,engineering, civil