Study On The Identification Of Uncertainty Parameter Distribution In Energy Consumption Prediction Of Urban Building Group

Urban building energy consumption accounts for more than 70% of the total energy consumption in the building sector,and there is a huge energy saving potential.Experts and scholars have pointed out that traditional energy-saving measures that only focus on the renovation of single buildings cannot cope with the increasingly severe challenges of energysaving and emission reduction.Detailed analysis and research on the characteristics and structure of urban-scale building energy consumption are required.Among the current urban energy consumption models,the most widely applicable and flexible one is the bottom-up randomness model based on physics.The model faces more severe challenges in calibration than the single building level.Among the randomness models,the Bayesian calibration technique is the most widely used.In previous studies,Bayesian calibration was only used to adjust the distribution of parameters at the building group level to match the energy consumption distribution output by the model and the actual energy consumption distribution of the building group.However,when the model is applied to the pre-evaluation of energysaving renovation measures at the building group level,the reliability of the parameter distribution is very important.This research is based on the method of simulation experiments,respectively studying: the prior distribution of hyperparameters involved in the widely used KOH Bayesian calibration framework;the method of using energy consumption data.The calibrated parameters identify the impact of performance,and verified the impact of the aforementioned two aspects in a residential area in Beijing.The results showed as follows: 1)the proposed regulation method of the hyper parameter prior distribution can significantly improve the energy consumption prediction performance of the model,and obtain the parameter posterior distribution closer to the actual distribution than the prior distribution;2)The iterative calibration method proposed in this study can significantly improve the identification of sensitive parameters and obtain the posterior distribution of parameters that can accurately represent the actual distribution.Finally,through a case of deterministic quantified energy conservation renovation measures,it is verified that the above methods can significantly improve the accuracy of energy consumption prediction after energy conservation renovation measures in the pre-evaluation of building groups.