都市建築環境における統計学手法の適用 : 健康－エネルギーの相互関係の分析

概要

The extensive use of energy in modern civilization is causing upheavals in the global climate, which is characterized by exceptionally high temperatures and humidity . Climate change with its adverse health effects is responsible for the intricate association between energy consumption and human health. A recent upsurge in climatic events, such as droughts, wildfires, and heatwaves, along with the global pandemic, has necessitated a subtle evaluation of the built environment, the preeminent contributor to global energy consumption, to cope with future unpredictable episodes. With this background, this thesis focuses on the built environment to address the mutual relationship between energy consumption and human health by applying advanced statistical analyses. A flexible generalization of the conventional regression model (ordinary linear regression model), namely the generalized linear model (GLM), was adopted as the analytical approach. The GLM relies on the link function (the function of the mean response) to model variant responses. Although the relationship between the response and predictor is not always linear, the link function allows the response to connect linearly with the predictors through GLM. As an umbrella term, GLM stands for a class of regression models to encompass different types of responses, such as continuous, binary (or categorical), and count.

This study initiated an analysis of the daily number of heatstroke transporters (HST) in hospitals of Fukuoka city, Japan, which is a count response. The 2 n d chapter of this thesis analysed these overdispersed count data (HST of Fukuoka city, Japan) through a negative binomial (NB) regression model, a variety of GLM. NB models included daily weather characteristics and modified factors as predictors to identify potential risk factors. Age - structure decomposition followed by the application of NB mode ls detected people aged 70 or more as the most vulnerable group for heatstroke. The potential determinants of heatstroke-risk were also investigated in susceptible people. The findings of the analysis highlighted the importance of preparing intensive care facilities and efficient preventive measures to withstand the adverse health impacts of climate change, specifically during extreme climatic events.

To mitigate global climate change, achieving energy sustainability is indispensable. A significant transition in energy consumption worldwide was encountered owing to the pandemic-induced restriction of human life and financial burden. The 3 r d chapter of this thesis investigated whether this energy transition was significant enough to practise energy-efficient lifestyle even in the post-pandemic era. Analytical procedures evaluated the energy consumption behavio ur of strictly restricted Malaysian people complying with the movement control order (MCO) during the pandemic period. The commonly used cooling applianc e of Malaysia, namely the air-conditioning system (AC), was targeted in the analysis. The ordinal logistic regression model analysed the AC -usage behaviour during MCO-day remote work with four ordinal categories. As predictors, the regression model investigated previous behaviour, dwelling and economic conditions along with demographic characteristics. Moreover, this study scrutinized the pandemic -driven behavioural changes in AC use to highlight the discernible shift toward energy conservation. The multinomial regression model analysed two types of behavioural changes like stopping and starting AC use during the MCO period along with the behaviour of no change. The findings indicated stopping AC use during remote work (using AC at the office before the MCO period but not during remote work in the MCO period) as the preeminent behavioural change. Potential factors of such behavioural change were decisive for strategy planners to design efficient programs and sustain efficient energy consumption behaviour.

To achieve energy sustainability exclusively, expeditious programs must be designed to encourage the efficient use of energy. To achieve carbon neutrality, the Japanese government increased the penetration of renewable energy generation and experienced an overgeneration of solar power in the Kyushu area. This situation necessitated the emphasis on the self-storage and consumption of rooftop solar photovoltaic panel owners. Thus, the electricity demand of single or smaller aggregated dwellings requires subtle analysis to secure the demand–supply balance. Chapter 4 analysed the electricity demand database from a residential building in Settsu city, Osaka, Japan, to identify stochastic features of peak demand from aggregated dwellings. The demand database from 479 dwellings was subsampled to create data of different aggregation sizes (3, 10, 50, and 479) and calculate the daily number of peaks. For aggregation sizes of 3, 10, and 50, the data on daily peak counts were obtained repeatedly for each subset and analy sed using generalized linear mixed models (GLMM) to accommodate intra -subset associations. However, the aggregated data on 479 dwellings had only one subset , and the GLM was applied to identify the stochastic features of peak demand. Hot and cold weather conditions were associated with a higher likelihood of peak consumption. Additionally, weekends or holidays with a day-long household presence exhibited higher peak consumption.

The significance of hot and cold weather conditions in increasing electricity consumption highlights the contribution of AC loads to the peak -demand acceleration. Thus, in chapter 5, the AC loads of each dwelling were scrutinized to evaluate their potentia l for demand response (DR) activities. A score analysis quantified the DR potential of target dwellings by investigating the behavio ural traits (frequency, consistency, peak -demand, and gridadversity) of AC use. As the data lacked information on individua l characteristics, the cluster analysis classified households based on energy consumption behaviour through specific appliances. Finally, quantile regression models, a robust alternative to linear regression models, were used to examine the association between DR potential scores and household clusters to determine potential household characteristics. Heavy ACconsumers and inhabitants of spacious dwellings were more suited to DR schemes. The day-long presence of at least one resident in the home was another important attribute of household recruitment in the DR programs.

As a whole, this thesis underlines the trilemma among energy, environment, and human health. In some selected critical sites, we investigated distinct perspectives to reveal the pathway toward an energy-efficient world in an effort to overcome the adverse health impacts of climate change.