Developing a Probabilistic Stock Turnover System Dynamics Model to Forecast Whole-life Energy of Chinese Urban Residential Building Stock
China is a major driving force of the growth of the global building stock – in 2018 alone, 2.5 billion m2 in new buildings were constructed, amounting to 34% of the global total. In the same year, the final operational energy consumption of Chinese buildings was 504 million tonnes of oil equivalent (toe). More than one third of this energy was consumed by urban residential buildings. With a growing urban population and mounting demand for energy services in the built environment, urban residential buildings play an increasingly important strategic role in China's efforts to decarbonise its building sector. Nevertheless, there is a basic lack of authoritative data on Chinese urban residential stock. Official statistics on total floor area of the stock only exist up to 2006. The historical stock growth trajectory from 2007 onwards is therefore unknown. This creates a major barrier to understanding and analysing future trends in stock evolution or energy use.
Meanwhile the embodied energy associated with the massive flow of materials consumed by the ongoing construction boom across China remains under-explored. Urban buildings in China are generally short-lived, implying a lower risk of operational energy lock-in as the stock is rapidly replenished with more efficient buildings; however, this comes at a cost of high embodied energy incurred by construction. Very limited studies have been conducted on the trade-offs between embodied and operational energy, resulting in little in the way of an evidence base to inform policymaking.
This thesis fills these key research gaps, by modelling both stock turnover dynamics and the whole-life (operational and embodied) energy of Chinese urban residential buildings over the medium to long term. Whilst the thesis excludes carbon from its scope, its outputs prepare the ground for quantifying energy-related carbon emissions of the stock.
A stock turnover model is developed using a System Dynamics methodology. The model captures the dynamic interplay between new construction, aging of existing buildings in operation, and demolition of old buildings. With survival analysis applied to building lifecycle, annual demolition of buildings is modelled as a stochastic process based on the hazard function of a Weibull distribution. Using official statistics on annual new construction over the historical period of 1978 to 2006, the parameters of the Weibull distribution are calibrated by fitting the stock turnover model to historical data on total floor area up to 2006. Based on the calibrated lifetime distribution and annual new construction data from 2007 onwards, the stock turnover model is run to estimate the total stock size of urban residential buildings and the dynamically changing age profile of buildings in the stock for each year up to the present day. One key result finds that the average building lifetime is 34.1 years, much shorter than the design lifetime of 50 years.
The stock turnover model is then extended to forecast future stock trajectories through Bayesian Model Averaging (BMA). In addition to the Weibull distribution, another four potential distributions for the survival model are considered, including Lognormal, Loglogistic, Gamma and Gumbel distributions. For each, the probabilistic stock turnover model is simulated using Markov Chain Monte Carlo (MCMC) methods. BMA is then applied to combine model-specific predictions of the historical stock evolution based on the respective probabilities of the five survival models. Finally, by extending model structure and incorporating variables relating to urbanisation and housing demand, future stock turnover dynamics and possible trajectories over the medium to long term are forecast. These results suggest that the floor area of urban residential stock in China is likely to peak around 2065, at between 42.4 and 50.1 billion m2.
The probabilistic stock turnover model and its results, which are core to the modelling and analysis of this thesis, lay the groundwork for estimating embodied energy, operational energy, and whole-life energy of the stock. The scope of embodied energy covers the different building lifecycle stages of material production, transportation, on-site construction, and demolition. Building materials are limited to steel, cement, aluminium and glass, the most commonly used for Chinese buildings. To assess operational energy, an improved SD model is developed which addresses some major methodological issues identified in previous SD-based models for operational energy. By adapting this generic model to the urban context in north China, where space heating requirements are most significant, and integrating it with the probabilistic stock turnover model, the operational energy is forecast. The results show that embodied and operational energy are likely to peak around 2027 and 2051, respectively.
Four realistic scenarios for the whole-life energy of urban residential stock in north China are then investigated. Under the business-as-usual (BAU) scenario, peak energy is expected to occur around 2050, with a mean value of 383.7 million tonnes of coal equivalent (tce). Extending building lifetime (Policy A) and accelerating heating energy efficiency improvement (Policy B) both reduce whole-life energy compared with the BAU scenario. Policy C integrates Policy A and Policy B and captures the endogenous dynamics of time-varying building lifetime distribution and heating energy intensities. Compared with the BAU scenario, Policy C achieves a cumulative energy saving of 35.7 million tce on average during the period 2025-2030. While this is under 10% of the total, it exceeds the sum of operational energy of buildings in 24 out of the total 31 provinces of mainland China in 2015.
The detailed findings of model development and application provide extensive evidence, based on which a series of policy recommendations are made. Amongst them, the most fundamental is that the trade-offs across stock-level embodied and operational energy should be analysed, and a whole-life perspective taken, in designing future policies for buildings in order to move towards China's target of carbon neutrality by 2060.
This thesis makes several important contributions. The modelled lifetime of 34.1 years quantifies and substantiates the observation that Chinese buildings are generally short-lived. The model development and its application are characterised by first-of-its-kind approaches to forecast stock evolution in a Bayesian framework, which is the major contribution of this thesis. This is also a first effort at integrating stock-level embodied and operational energy in a single model, in order to draw conclusions about whole-life energy for China’s urban residential building stock under a number of policy scenarios. The quantified whole-life energy, along with the characterisation of building stock turnover dynamics and the probabilistic forecasting of future stock evolution, will enable further analysis to develop insights into whole-life carbon emissions of the building stock, which is of strategic importance under China’s announced climate targets. Methodologically, with high generality and flexibility, the model can now be adapted to a wide variety of geographical contexts.