Machine learning-based analysis of occupant-centric aspects: Critical elements in the energy consumption of residential buildings

https://doi.org/10.1016/j.jobe.2021.103846Get rights and content

Highlights

  • A framework that couples machine learning techniques with orthogonal experiment design is developed.

  • Occupant-related parameters criticality is tested with building envelope parameters.

  • A case study is conducted to simulate building energy consumption.

  • Cross comparison of the framework outcomes with literature findings is discussed.

Abstract

The housing sector consumes a significant amount of energy worldwide, which is mainly attributed to operating energy systems for the provision of thermally comfortable indoor environments. Although the literature in this field has focused on investigating critical factors in energy consumption, only a few studies have conducted a quantitative sensitivity analysis for thermal occupant factors (TOF) (i.e., metabolic rate and clothing level). Therefore, this paper introduces a framework for testing the criticality of TOF with a cross-comparison against building-related factors, considering the constraint of occupant thermal comfort. Using a building energy simulation model, the energy consumption of a case study is simulated, and building energy model alternatives are generated. The scope includes TOF and building envelope factors, with an established orthogonal experimental design. A popular branch of machine learning (ML) called linear genetic programming (LGP) is used to analyse the generated data from the experiment. Finally, a sensitivity analysis is conducted using the developed LGP model to determine and rank the criticality of the considered factors. The findings reveal that occupants' metabolic rate and clothing level have relevancy factors of −0.48 and −0.38 respectively, which ranked them 2nd and 3rd against building envelope factors for achieving energy-efficient comfortable houses. This research contributes to the literature by introducing a framework that couples orthogonal experiment design with ML techniques to quantify the criticality of TOF and rank them against building-envelope factors.

Introduction

The housing sector consumes one-third of the global energy; yet, its performance has not been a success [1]. The basic aim of houses is to offer safe and comfortable private environments, which they do, provided the occupants remain able to afford the operational costs. Unfortunately, such costs are not only economic but also environmental in our life. This has driven the industry towards energy-efficient solutions to mitigate impact on the environment by minimising greenhouse gas emissions and energy consumption [2]. Thereby, occupants are put in a position where they must sacrifice their comfort to save energy [3]. This approach has minimised energy consumption but makes occupants feel uncomfortable in their dwellings.

A building design can play a major role in exploiting outdoor environments to passively provide a comfortable indoor environment [4]; it can also help building energy systems achieve minimal operation. This requires minimal design constraints or even their absence. However, most designs are constrained due to many reasons, including municipal wellbeing requirements. Hence, passive designs may sound potentially promising but are limited in terms of application.

One of the most critical factors in building energy consumption is the building envelope [5]. Its design differs between buildings due to the different requirements, including building location, weather behaviour, and occupants' needs. For example, a building with a full glass façade would take in more daylight, causing high indoor radiant temperature. This would be thermally uncomfortable for the occupants during warm seasons and would require the heating, ventilation, and air-conditioning (HVAC) systems to work more than it would if the façade had only windows. This might contradict with the fact that a modern building must be architecturally aesthetic. Therefore, building energy simulation (BES) has been one of the main research streams in the building industry in recent years [6]. This is attributed to the multidimensional impact of energy consumption on the building sector, including environmental emissions, economic potentials, and the lifestyles of residents [7]. Therefore, a finer level of research has come to researchers' interest, investigating critical parameters that buildings' energy behaviour is mostly sensitive to Refs. [[8], [9], [10], [11], [12]]. It has been shown that building envelope parameters have different impacts on the overall building energy consumption (BEC) [13]. Additionally, Li et al. [14] performed a study to investigate the most sensitive design parameters of buildings in sub-tropical regions using three different sensitivity analysis (SA) techniques.

In addition to building energy consumption, occupant thermal comfort is an important criterion to assess the building envelope performance. Thermal comfort is necessary for humans to experience an efficient and healthy lifestyle [15]. HVAC systems have filled the thermal comfort gap that differentiates indoor from outdoor environments using energy. Therefore, a contradiction between energy efficiency and improved occupant thermal comfort exists [16]. The amount of energy required to amend indoor thermal comfort level is dependent on several factors such as building envelope and occupant characteristics [17,18]. In general, a proper and efficient design of a building involves optimised occupant comfort levels with lower amounts of energy consumption [19]. Hence, the energy consumption of buildings should be simulated with occupant comfort in mind, either as an objective or as a constraint [20]. Bünning et al. [21] emphasised this concept by conducting an experimental demonstration through data predictive control to optimise the operation of energy systems for both occupants' thermal comfort and energy performance. This continuous style of optimisation is important since Lundqvist et al. [22] showed that the energy-retrofit of households may be inefficient in cutting operational costs if the thermostat is not adjusted to provide a comfortable environment.

Occupant thermal comfort can be used to make HVAC systems work accordingly, ensuring that it is a constraint in the process rather than an objective [23]. This approach allows the researcher to focus their attention on energy consumption and building-occupant factors. One of the most popular methods to assess thermal comfort is the predicted mean vote (PMV) index, it is widely used in literature for such a purpose [24]. Wu et al. [23] created a model for an office building and compared two strategies of controlling the HVAC system: air temperature-based control and PMV-based control. Their results showed that the PMV-based control model improves the occupant thermal comfort significantly, in addition to 1.6% savings in energy consumption, compared to the air temperature-based control model.

PMV-based control strategy has not been widely applied to detached houses due to different reasons such as occupants' desire to control their HVAC settings [25]. Occupants tend to have flexible and ready-to-adjust HVAC systems, especially if their lifestyle involves non-routine activities. This is attributed to the number of variables in the PMV model that need to be assumed such as occupant's mechanical work, and indoor air velocity [26]. Therefore, uncertainty leads to uncomfortable PMV-based settings mostly due to thermal occupant factors (TOF).

The existing practice of the building industry is to focus on the building physics rather than the occupants' characteristics due to many reasons such as ambiguity of occupant's nature, and avoiding bias towards specific occupant's needs if bulk housing projects are being designed [27]. After exploring the literature, while building envelope parameters have been studied extensively, little work has been done to investigate a quantifiable criticality of TOF. Therefore, the current research focuses on addressing the identified gap in the literature.

TOF such as metabolic rate (Met) and clothing level (Clo), are local factors playing a role in the process of indoor thermal comfort. Identifying critical factors in the building-occupant environment helps designers improve their products in terms of energy consumption by focusing detailed attention on them and optimising those factors. Since the impact of building envelope factors has been explored in this manner, the literature lacks studies that investigate a quantifiable criticality of TOF and how they compare to building envelope factors. In this context, the presented research is performed to answer the following questions:

  • -

    What are the most critical building envelope and TOF factors in terms of energy consumption?

  • -

    How do PMV-based HVAC control settings change building energy consumption compared to typical settings?

  • -

    How significant is the impact of TOF on the objective function of energy consumption?

In this research, a linear genetic programming (LGP) model is developed using BEC-based orthogonal experiment data, which are to be simulated in an energy simulation engine using a dwelling case study. The contribution of this study lies in introducing a framework that involves ML techniques to investigate the criticality of TOF in the context of BEC. This approach brings attention to the importance of familiarising designers with occupants' thermal characteristics beforehand.

The remainder of the presented paper is structured as follows: Section 2 gives a brief background of energy simulation and critical building factors, along with the gaps in the literature. Section 3 highlights the framework of the method and the case study used to perform this research, and Section 4 discusses the results in terms of BEC, orthogonal experiment, and the LGP. Furthermore, a detailed discussion of the results and inferences, following the same categorisation of Section 4, is presented in Section 5. Finally, Section 6 presents the conclusion of this research.

Section snippets

Framework

The framework of this research is carefully designed (as shown in Fig. 1) to answer the research questions. More explanations of different phases of this study are given in the following.

  • -

    The framework starts with the BES phase (left-side part of Fig. 1), in which a building energy model is developed in an energy simulation environment to test the functionality of the model for generating total energy consumption results. To develop this model, detailed inputs must be made, as will be

Building energy simulation

The initial values of the parameters are made for the base case to carry out BEC simulation as shown in Table 2, namely, Met, Clo, EWTR, window thermal conductivity (WTC), building orientation (BO), infiltration rate (Inf), GSTR, and WSA. The total consumed energy for one year is 8760.88 kWh.

Fig. 5 shows the results for the base case, including outdoor temperature, indoor air temperature, and occupant thermal comfort. It can be noticed that the outdoor temperature fluctuates between 2 °C and

Building energy simulation

The results of the base case simulation suggest 8760.88 kWh, which is comparable to the annual average energy consumption of a household in Victoria – 9397 kWh [44]. The low energy consumption result, despite the occupant being comfortable, can be explained by the comfort thresholds in the BES model. The International Organization for Standardization (ISO) suggests a PMV range between −0.5 and +0.5 for optimal thermal comfort [45], whereas a range of −1 to +1 is used in this study. This gives

Conclusion

This research conducts a criticality analysis using orthogonal experiment and linear genetic programming (LGP) techniques to identify the criticality ranking of building parameters, including building envelope factors and thermal occupant factors (TOF). The framework of this research comprises three main phases i.e., building energy simulation (BES) phase, design of experiment phase, and machine learning phase.

Upon observing the statistical indices, the LGP model performed well in generating

Author statement

Alsharif, R., Arashpour, M., Golafshani, E., Chang, V. and Zhou, J. conceived and planned the experiments. Alsharif, R., Arashpour, M. and Golafshani, E. carried out the experiments. Alsharif, R. and Golafshani, E. planned and carried out the simulations. Alsharif, R. and Arashpour, M. contributed to sample preparation. Alsharif, R., Arashpour, M., Golafshani, E., Chang, V., Zhou, J. and Hosseini, M. contributed to the interpretation of the results. Alsharif, R took the lead in writing the

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (53)

  • K.-U. Ahn et al.

    Correlation between occupants and energy consumption

    Energy Build.

    (2016)
  • Z. Xu et al.

    PMV-based event-triggered mechanism for building energy management under uncertainties

    Energy Build.

    (2017)
  • L. Deng et al.

    An optimization method for multi-objective and multi-factor designing of a ceramic slurry: combining orthogonal experimental design with artificial neural networks

    Ceram. Int.

    (2018)
  • M. Castelli et al.

    Prediction of energy performance of residential buildings: a genetic programming approach

    Energy Build.

    (2015)
  • M. Arashpour et al.

    Scene understanding in construction and buildings using image processing methods: a comprehensive review and a case study

    J. Build. Eng.

    (2021)
  • A. Zhang et al.

    An integrated school and schoolyard design method for summer thermal comfort and energy efficiency in Northern China

    Build. Environ.

    (2017)
  • N.F. Shur et al.

    Age-related changes in muscle architecture and metabolism in humans: the likely contribution of physical inactivity to age-related functional decline

    Ageing Res. Rev.

    (2021)
  • T. Chaudhuri et al.

    A feedforward neural network based indoor-climate control framework for thermal comfort and energy saving in buildings

    Appl. Energy

    (2019)
  • B. Chen et al.

    Multiobjective optimization of building energy consumption based on BIM-DB and LSSVM-NSGA-II

    J. Clean. Prod.

    (2021)
  • H. Li et al.

    Solar energy absorption effect of buildings in hot summer and cold winter climate zone, China

    Sol. Energy

    (2020)
  • E. Cuce

    Accurate and reliable U-value assessment of argon-filled double glazed windows: a numerical and experimental investigation

    Energy Build.

    (2018)
  • X. Chen et al.

    A comprehensive review on passive design approaches in green building rating tools

    Renew. Sustain. Energy Rev.

    (2015)
  • A. Ioannou et al.

    Energy performance and comfort in residential buildings: sensitivity for building parameters and occupancy

    Energy Build.

    (2015)
  • G.J. McRae et al.

    Global sensitivity analysis—a computational implementation of the Fourier amplitude sensitivity test (FAST)

    Comput. Chem. Eng.

    (1982)
  • J. Nossent et al.

    ‘Sobol’ sensitivity analysis of a complex environmental model’

    Environ. Model. Software

    (2011)
  • R. Aghamolaei et al.

    Balancing the impacts of energy efficiency strategies on comfort quality of interior places: application of optimization algorithms in domestic housing

    J. Build. Eng.

    (2020)
  • Cited by (13)

    • Prediction of wildfire rate of spread in grasslands using machine learning methods

      2022, Environmental Modelling and Software
      Citation Excerpt :

      ML uses input data for the purpose of learning and then predicts new situations based on what was learned. Improvement in data gathering procedures and computational power has opened the range of applications for ML models (Alsharif et al., 2022). ML techniques are being utilized widely for developing predictive models in environmental applications (Zumwald et al., 2021).

    View all citing articles on Scopus
    View full text