Characterising Chinese Urban Residential Stock Turnover Dynamics using Bayesian Model Averaging
讲座名称：
Characterising Chinese Urban Residential Stock Turnover Dynamics using Bayesian Model Averaging
讲座时间：
2019-10-19
讲座人：
周维
形式：
校区：
兴庆校区
实践学分：
讲座内容：
报告题目：Characterising Chinese Urban Residential Stock Turnover Dynamics using Bayesian Model Averagi
报告时间：2019年10月19日，9：30-10：30
报告地点：教工之家会议室
报告人：周维
报告摘要：
Building stock turnover is a key determinant in building energy modelling. In turn, building lifetime is integral to the dynamics of stock turnover. The building stock of China, as evidence for building energy policies, is a strategically important but under-researched area. Despite anecdotal claims that urban residential buildings are generally short-lived, there are no official statistics on building lifetime, and empirical data is extremely limited. Moreover, official statistics on total floor area of urban residential stock in China only exist up to 2006. Previous studies estimating Chinese urban residential stock size and energy use have made various questionable methodological assumptions and only produced deterministic results. This paper presents a Bayesian approach to characterise the stock turnover dynamics and estimate stock size uncertainties for the period from 2007 to 2017. Firstly, a probabilistic dynamic stock turnover model is developed to describe the building aging and demolition process governed by a hazard function specified by a parametric survival model. Secondly, with each of five candidate parametric survival models and using official statistics up to 2006, the dynamic stock turnover model is simulated through Markov Chain Monte Carlo (MCMC) to obtain posterior distributions of model-specific parameters, estimate marginal likelihood, and make predictions of stock size. Finally, Bayesian Model Averaging (BMA) is applied to create a model ensemble that combines the model-specific posterior predictive distributions of the 2007-2017 stock evolution pathway in proportion to posterior model probabilities. The distribution of building lifetime, unconditional on the survival models and the model-specific parameters, is obtained. This study is a first-of-its-kind use of a full Bayesian approach to investigate model and parameter uncertainties that were not taken account of by limited existing models targeting Chinese building stock. The Bayesian modelling approach and the results can serve as a baseline for further studies on forecasting building stock development trajectory and analysing energy and carbon impacts. This will have particular relevance for modelling and analysing policy scenarios to investigate the trade-offs across embodied-versus-operational energy and carbon emissions facing Chinese residential buildings. This information will be critical for sector-wide transformation towards low-carbon buildings, as the Chinese Government pledges to peak its overall emissions by 2030.
相关视频