Indoor environmental quality evaluation of smart/artificial intelligence techniques in buildings – a review

. The built environment sector is responsible for around one-third of the world's final energy consumption. Smart technologies play an essential role in strengthening existing regulations and facilitating energy efficiency targets. Smart Buildings allow the response to the external conditions of buildings including grid and climatic conditions, and internal building needs such as user requirements achieved through real-time monitoring and real-time interaction which are resembled the smart buildings concept. The optimal management of occupant comfort plays a crucial role in the built environment since the occupant's productivity and health are highly influenced by Indoor Environmental Quality. This work explores the application of real-time monitoring and interaction to achieve optimal Indoor Environmental Quality, occupant comfort and energy savings in relation to smart buildings and smart technologies. To better address and indoor air quality issues, ventilation needs to become smarter. It is crucial to understand first the Key Performance Indicators of evaluating smart ventilation. In parallel, Artificial Intelligence techniques such as machine and deep learning have been increasingly and successfully applied to develop solutions for the built environment. Thus, this paper provides a review on the existing Key Performance Indicators that allows smart ventilation in smart buildings. Then, it reviews the existing literature on the machine and deep learning methods and software for assessing the smart ventilation. Finally, it shows the most recent technologies for performing experimental evaluation on the main indicators for smart ventilation. This work is expected to highlight the selection of the most optimal ventilation metrics, proper indicators, machine learning and deep learning models and measurement technologies to achieve excellent Indoor Environmental Quality and energy efficiency levels.


Introduction
Buildings in the European Union (EU) are responsible for 40% of the total energy consumption and 36% of the European global CO2 emissions [1]. The most significant proportion of building energy consumption goes for heating and cooling, accounting for 70% of the energy consumption of residential building stocks [2]. The energy consumed in buildings provide the occupants with healthy and comfortable indoor environments to live and work in, as they spend more than 90% of their time indoors. Several energy targets and requirements have been developed to overcome these issues and achieve carbon reduction and energy efficiency in buildings. With the shift towards the smarter grid system and smart metering in buildings and the ongoing technological advancements, the concept of Smart Buildings (SBs) has been established [3]. Smart Buildings has been defined in [4] as "A nearly Zero Energy Building (nZEB) that can manage the amount of Renewable Energy Sources in the building and the Smart Grid through advanced control systems, smart meters, energy storage, and demand-side The presence of smart technologies has made it possible to tackle several issues related to the IEQ, building energy and occupant comfort and satisfaction. Several aspects need to be considered for proper management of energy, comfort, and air quality. Recently, there has been a significant increase in the development of SB control systems to increase indoor air quality by connecting monitored environment variables (e.g., temperature, humidity, luminosity, and air quality) to building management systems (e.g., heating, ventilation, and airconditioning (HVAC) system, lighting system) [7]. Moreover, Artificial intelligence (AI) techniques such as machine and deep learning have been increasingly and successfully applied for managing the building energy efficiency, thermal comfort, and air quality [8]. Furthermore, it is vital to understand the Key Performance Indicators (KPIs) for managing and evaluating the indoor air quality, and occupant comfort in SBs. KPIs are quantifiable measurements, that provide a framework or a set of good practices that should then be followed within the operation of the building [9].
Thus, this paper gives a review on the application of real-time monitoring and interaction in SBs to achieve optimal IEQ, occupant comfort and energy savings in relation to smart technologies with special reference to holistic approaches.

Methodology
The aim of this research is to understand and define the appropriate smart technologies for assessing the occupants' comfort and IEQ. Thus, this paper provides a holistic review on the technologies and indicators that evaluates smart technologies utilization in SBs to allow user interaction and achieve optimal indoor air quality. In this paper, the literature review is sorted into numerical and experimental parts as presented in Figure 1. The numerical literature review represents the state-of-the-art literature review in three main areas; 1) smart technologies evaluation for occupant comfort, health, satisfaction and IEQ, 2) selection and application of appropriate advanced control strategies, AI, and machine learning for managing and evaluating user interaction and IEQ, 3) the main influencing KPIs that should be considered when evaluating the user comfort and IEQ. Then the experimental literature review focuses on exploring the in-site equipment that test the occupant comfort in SBs. The methodology adopted in this article is based on a structured systematic literature review. It aims at addressing several questions including: What are the possible control strategies and methods to study the occupant behaviour, comfort and allow maintain proper IEQ? What are the most influencing KPIs and metrics to study the occupant behaviour in buildings? What are the current smart technologies used to evaluate the occupant comfort and measure the IEQ in SBs?

State-of-the-art
Building's occupants can directly affect the energy consumption in the built environment. Knowing the occupancy information in the building environment is an important parameter for efficient energy use. In order to understand the occupant behaviour pattern in buildings, different types of smart technologies have been applied and studied in literature. Moreover, Occupant health and productivity in buildings depend on the IEQ [10]. Clearly defined occupancy information in buildings can help to improve thermal comfort, user satisfaction, energy saving, and indoor air quality [11]. Building technologies can be harnessed by deploying sensors inside buildings, to collect relevant data about both energy consumed and occupant behaviour, since occupants influence building appliances, such as HVAC, lights, and hot water tanks. The next sections investigate the application of smart technologies and their relationship with the IEQ. Moreover, it gives an insight for the main indicators used to investigate the IEQ in SBs.

Smart
Technologies for evaluating occupants' comfort and IEQ in Buildings SBs solutions rely on a combination of well-established and emerging information and communication technologies (ICT), such as wireless networks, Internet of Things (IoT) and cloud computing. One of the most discussed smart technological solutions are the sensors. Building sensing technologies have progressed rapidly in the last two decades in aid of monitoring IEQ and energy system performance. Analysing the occupant behaviour using occupancy sensors helps to enhance energy saving in the built environment [12]. Several kinds of sensors are used to measure the indoor parameters of the building, such as the temperature, luminosity, occupancy, and energy consumption [13]. In [14], sensors for smart buildings operations were classified as three categories including; Occupancy sensors (e.g. Image based sensor, Passive infrared (PIR) sensor, radio-based sensor, Photo sensor, Ultrasonic doppler, Microwave doppler, Ultrasonic ranging), Built environment measurements (e.g. CO2 sensor, Air Temperature sensor, Humidity sensor, Thermo-fluidic sensor, Sound sensor, Light sensor, Volatile organic compound sensor, Particulate Matter (PM) sensor, Air velocity sensor) and other sensors (e.g. IoT based sensor, Smart Phones, Heart Rate sensor, Fingerprint sensor, Mobile pupilometer, Skin Temperature Sensor). The application of these sensors has been widely implemented for building energy and comfort management (BECM). where the objective of BECM is to optimize, analyse and control energy consumption, and achieve better IEQ [15]. Moreover, heating, ventilation, and air conditioning (HVAC) smart systems use multiple sensors for monitoring and control. Sensors used in HVAC systems are used to monitor temperature (e.g., outside air temperature, chilled water temperature, supply air temperature), humidity, flow (e.g., chilled water flow rate, supply airflow rate), pressure (e.g., chiller water pressure, duct static pressure), and gas flow for absorption chillers or boilers [16]. Occupants' presence information can be used to turn on/off lighting and HVAC systems, and adjust temperature set points. While the number of occupants can be used to adjust ventilation. Moreover, occupant identity and location can be used to provide customized indoor environments to maximize their satisfaction while optimizing energy efficiency [17]. Smart HVAC systems use multiple sensors for monitoring and control. Software interprets information from various sensor points to optimize the HVAC system's operation while improving occupant comfort. Smart HVAC controls can limit energy consumption in unoccupied building zones, detect and diagnose faults, and reduce HVAC usage, particularly during times of peak energy demand.
On the other hand, the concept of virtual sensing was discussed as an alternative for physical sensing. Virtual sensing has been used for SB applications, in particular, for HVAC systems monitoring and fault diagnostics [18], [19]. In [20], a virtual occupancy sensor was presented for real-time occupancy information in buildings using a Bayesian belief network algorithm to fuse data based on real-time Global Positioning System (GPS) location and Wi-Fi connection from smart devices such as smart phones and occupancy access information from BEMS. In [21], the occupancy presence using the trajectory of indoor climate sensor data was identified by applying a set of rules to determine a probability of occupancy. However, the virtual sensing technology has uncertainty that is associated with physical sensor errors. Another technology that has evolved towards intelligence, and usability to achieve a better control and an improved energy performance in SBs is the thermostat. Smart thermostats were developed to remove the barrier of programmed users' schedules of the thermostats by automating this control as much as it can [22]. In [23], it was claimed that smart thermostats cover five areas, i.e., working as a basic thermostat in the absence of the connectivity, supporting the schedule-based HVAC operation, providing customers with energy feedbacks on the thermostat settings, providing the information about energy usage, and working with utility programs to prevent brownouts and blackouts. Figure 2 summarizes the main smart technologies that can be implemented in SBs ranging from sensors, virtual sensors, smart thermostats, IoTs, cloud system advanced HVAC system, control system and BEMS to assess and evaluate the IAQ and occupant behaviour.

Advanced control Strategies and BEMS for User Interaction and IEQ in Smart Buildings
Occupancy-based building system control is defined as a control method that adjusts the building system operation schedules and setpoints based on the measured occupant behaviour and has been identified as a smart building control strategy that can improve building energy efficiency as well as occupant comfort [24]. The goals of intelligent control/management systems for the IEQ and comfort includes ; high comfort level (to learn the comfort zone from the user's preference, and guarantee a high comfort level), energy savings through combining the comfort conditions control with an energy saving strategy and air quality control to Provide CO2-based demandcontrolled ventilation (DCV) systems [25]. Different approaches for controlling indoor building environments have been developed to adapt to SBs and allow occupant monitoring which are called computational intelligence technologies [26]. Artificial intelligence control systems are optimized by the use of evolutionary algorithms and developed for the control of subsystems of an intelligent building. The following sections shows the available advanced control systems and BEMS that are used for evaluating IEQ in SBs.

Fuzzy systems
Fuzzy and neural controls are the control techniques used for occupant comfort and energy efficiency. Their practical applications for HVAC systems have been applied widely, aiming for performance improvement over classical control [27]. The main obstacle to the application of traditional control methods in buildings is the mathematical model of the building operation.
Integrating new-type, higher-level variables that define comfort into the intelligent and advanced controllers, it is possible to control comfort without going into the regulation of lower-level variables like temperature, humidity, and air speed. This would allow user interaction and the participation in the specification of the desired comfort. Genetic Algorithms and methods coming from the theory of adaptive control are used to optimize fuzzy controllers. Fuzzy logic control systems had been used in a new generation of furnace controllers that apply adaptive heating control to maximize both energy efficiency and comfort in a private home heating system [28]. Fuzzy controllers were widely used to control thermal comfort, visual comfort, and natural ventilation and has led to significant results [29]. The inputs of a fuzzy controller are Predicted Mean Vote (PMV) and outdoor temperature. Auxiliary heating (AH), auxiliary cooling (AC), and ventilation window opening angle (AW) settings are the controller outputs [30]. The Fuzzy PID controllers are classified into two major categories, according to their structure [31].

Artificial Neural Networks (ANN)
Machine learning techniques for forecasting is a part of artificial intelligence where algorithms learn from data. Machine learning models can include artificial neural networks (ANNs), deep learning, association rules, decision trees, reinforcement learning, and Bayesian networks [32]. ANN are broadly applied to forecast building energy consumption [33], heating and cooling loads [34], control of HVAC Systems [35], and comfort enhancing [36]. An ANN model uses machine learning methods to learn a particular relationship between input and output parameters, and it identifies the relationship after being trained with sufficient input and output information [37]. ANNs are self-learning systems that can adjust their approach and adapt to variable environments when processing many types of information [38]. They are important tools for energy behaviour modelling in buildings. ANN has been tested in several studies to predict ventilation performance in buildings [39]. The main issue in ANN requires establishing a data-driven mathematical model for the prediction. ANNs have been widely used in pattern recognition to forecast changes, predict performances, improve accuracy, optimize decision making, and behavioural modelling. One of the main advantages ANN algorithms is that they allow memorizing training information and self-learning, optimizing of information and knowledge factors that impact the testing results [41]. Selfadaptability is the most critical aspect of ANN algorithms which influences the results' accuracy compared to conventional algorithms [42].

Model Predictive Control
Model predictive controller (MPC) is a method to design structures of control inputs to optimize an objective considering defined and driven constraints [43]. The controller uses the modelled system, system inputs, and forcing factors (e.g. outdoor weather, occupancy, solar gain, etc) to predict future conditions (e.g. indoor temperature) to make the most efficient control action [44]. MPC strategies are widely known for their efficiency and robustness, as they incorporate weather conditions, occupancy, building physics, etc., to predict the desired indoor setpoints [45]. Its main advantage is considering the future forecast of outdoor temperature, solar radiation and occupancy rate, equipment, weather, and cost in the design of control system which eventually provides the required level of thermal comfort [46].
A common application of MPC in the building sector comprises the prediction of the dynamic behaviour of systems in the future and adjustment of response by the controller accordingly leading to energy and cost saving while satisfying thermal comfort [47]. Another feature of MPC is the consideration of the occupant preferences as a direct input into the control algorithm, where occupants can report their perceived comfort. The incorporation of the occupant preferences reported through feedback in the control systems was previously elaborated [17]. It was claimed in several research that MPC utilization to control HVAC system or room temperature can save 10%-50% energy depending on controller configuration, and disturbances predictions [48]. Furthermore, several functions are associated for MPCs in SBs and can be summarized as follows [49]: Weather prediction and response: MPC can respond to climate conditions (weather forecast) and apply passive/active measures to maximize energy efficiency and minimize the energy taken/fed into the grid User prediction and response: MPC can learn from occupant's behaviour and impact of occupants on internal gains/loads estimation. It can switch off the technical systems in unoccupied periods and guarantee the thermal comfort of users in the occupied period [50]. Grid predictions and response: when dynamic electricity prices are applied to the system, MPC defines a load scheduling for the system to regulate correlation between time of consumption and peak load shifting/shaving/matching with the grid thereby energy/cost saving [51]. Thermal mass prediction and adaptation: MPC can use the potential of the building thermal mass to modulate the energy generation, consumption and storage of the system [52]. It can use the building thermal mass to shift energy demands to off-peak hours through adaption mechanisms.

User Interface
Smart buildings offer a good opportunity for occupant interaction through user interface. In SBs, the building automation and control systems offers user interface devices such as thermostat, valves, keypads, screens, mobile applications, etc., so users can interact with the components of each function (lighting, daylighting, heating, cooling, ventilation, shading, security). The system allows users by setting their preferences (desired comfort conditions, energy management, and occupancy schedule). In [53], the user interface was defined as anything the occupant can interact with within a building and may affect building services, energy use or internal environmental conditions. Several factors drive occupants to interact with the built environment. They can be divided into internal and external drivers [53]. Internal drivers can include age or body composition, physiological response to the environment, preferences, and habits. While external drivers include building characteristics, and building type, technological innovation, cost of electricity and availability of smart technologies. Moreover, occupants can adjust an interface for their own benefit and to meet their needs [54]. Some research showed the implementation of the user interface in SBs. In [55], a smart card unit (kiosk) was used which performed as the interface between the system and the occupants. Occupants' preferences are monitored via the smart card unit and considers the users' preferences for a specific time, such as one week. A statistical analysis is then achieved evaluating the users' preferences concerning three indoor comfort-controlled variables including PMV index, indoor illuminance, and CO2 concentration. In [56], a user-interactive occupants interface was developed that adjusted their temperature setpoints while providing feedback on their comfort preferences and received energy use information that was real-time, and directly related to specific actions. The interface was connected to an energy efficient model predictive HVAC controller. A user interactive prototype was implemented in [57] with an indoor environment control interface that shared HVAC energy use information with users while assembling their feedback on thermal preferences.

Key Performance Indicators Evaluating Occupant Comfort and IEQ in Buildings
The need to identify the Key Performance Indicators (KPIs) for the IEQ and occupant comfort is crucial for measuring the performance of existing implemented strategies and their future improvement. According to [58], KPIs are a way of measuring the performance of an organization and its success in achieving goals. In [59] it was claimed that indicator systems can provide measurements of the current performance and give a clear view of achievement in terms of future performance targets and progress. KPIs allows a uniform framework to quickly analyse and benchmark the performance for these characteristics in a given context. Several factors and indicators can have an influence on the Indoor air quality and the occupant comfort. In this regard, KPIs will identify target values to be set for the sensors, actuator connection, monitoring and user interfaces. Several studies have identified the different KPIs related to these issues. However, these KPIs are usually related to certain criteria and categories within the building. Several categories for the indicators were identified. For instance Brown [60], claimed that IEQ indicators included four factors that were: indoor air pollutant levels, thermal comfort, lighting and noise. While in [42] the ventilation performance and IAQ were found to be influenced by several factors and indicators such as: temperature and humidity ratio, windows and doors opening ratio, outdoor wind speed, and airflow ratio. Whereas [61], the indicators were related to thermal comfort, and included Predicted Mean Vote (PMV) and Predicted Percentage of Dissatisfied (PPD), as well as energy consumption indicator. Thus, in this paper, we categorized the most relevant and studied KPIs for measuring the IEQ and occupant comfort in Table 1. The most common categories that identify the KPIs for quantifying the IEQ are the: visual comfort, indoor air quality, and thermal comfort It represents the KPIs, their definition, category, and relevant references. The percentage of time that daylight levels are above a specified target illuminance.

Discomfort Index
The absolute distance of the observed value to the defined optimum value (22°C with 45% relative humidity) relative to the present comfort range (3°C and 10% relative humidity deltas).

Mean Vote (PMV)
It predicts the mean response of a large group of people according to the ASHRAE thermal sensation scale, on a scale from very cold (−3) to very hot (+3).

Percentage of Dissatisfied (PPD) Index
Percentage of the people who felt more than slightly warm or slightly cold for a given set of environmental conditions. Thermal comfort [66] Long-Term

Percentage of Dissatisfied (LDP)
A weighted average of the possibility of a thermal discomfort felt by the occupants of a space, considering the occupancy rate of the space and the time interval for which thermal comfort is calculated at one moment.

Air Quality Index (AQI)
A score with a range from 0 to 500 used to evaluate the air quality considering five major pollutants.

Experimental Equipment to Measure Occupant Comfort
Building occupant's patterns have significant effect on the overall energy performance of buildings [69]. The main reason of energy use is to ensure a comfortable indoor environment in terms of thermal comfort, visual comfort acoustic comfort and indoor air quality [70]. One of the difficulties of assessing comfort is that occupants' satisfaction and comfort are subjective variables. Several methods are reported in the literature to model user behaviour in built environments. However, previous research showed a gap between simulation results and in situ measurements. In parallel to modelling approaches, occupants' comfort can be explored and evaluated by means of experiments and subjective feedback like Thermal Sensation Vote (TSV), Thermal Comfort Vote (TCV), and Thermal Preference (TP). Previous studies have shown that the TP and TSV are the most widely used methods followed by the TCV [71]. The occupant comfort feedback can be collected by different ways. A vast number of studies used web tools and smartphone apps to collect occupant feedbacks [71]. Experiments can help to determine direct parameters (CO2 concentration, globe temperature, air velocity, relative humidity, and air temperature) or indirect data by means of interviews with occupants to determine their occupant satisfaction and patterns and individual information [72]. Figure 3 shows the black globe temperature sensor that provide the mean radiant temperature. It permits to calculate the operative temperature, which gives an indication on the thermal comfort. The operative temperature can be determined by means of the mean radiant and ambient air temperatures and the heat transfer coefficients. To maximize comfort while saving energy, occupant comfort control strategies have been investigated using occupancy and indoor environment, and outdoor climate data to control the building HVAC system. Several occupancy sensing devices were developed to detect the presence and behaviour in the indoor space and control the building system operations. The occupancy sensors are generally split into four types: Motion sensors, Imagebased sensors, Radio-based sensors and threshold and mechanical sensors [24]. Motion sensors, image-based sensors, radio-based sensors, and mechanical and threshold sensors (1). The most used sensors are the motion ones such as passive infrared (PIR). They can be used to determine the heating, cooling, and lighting scenarios. These sensors have a low accuracy. Imagebased sensors can provide information related to occupant's presence and number. The sensors might include infrared camera sensors, visible light camera sensors, and luminance camera sensors. Accuracy is higher for threshold and mechanical sensors (swiping cards to enter, stepping across a piezoelectric mat, etc.,) when it comes to occupancy count. The energy use can indicate a certain pattern of the interaction between occupants and the indoor environment. Figure 4 represents an example of sensors for energy consumption and consumption of DHW.

Discussions and Conclusions
There has been an increasing interest in real-time monitoring of data to ensure comfortable conditions. This literature review aimed at exploring the application of smart technologies in SBs to achieve optimal IEQ, occupant comfort as well as energy savings. This review paper helps in facilitating the selection of appropriate advanced control strategies, smart technologies for SBs to manage and evaluate user interaction and achieve proper IEQ. The paper reviewed the state-of-the-art of the smart technologies used to evaluate the occupant behaviour, comfort, and IEQ. The review showed that the most promising technologies includes the sensors/virtual sensors, smart thermostat systems, cloud systems and advanced control systems as well as smart HVAC systems. These support monitoring and controlling the IEQ. ICT-related applications can provide a simplified communication between users within indoor environment. The review showed that users can take advantage of smart technologies implemented in SBs. In order to allow this, users need to monitor and control both outdoor and indoor environmental parameters in realtime. To provide an efficient real-time energy use feedback, it is important to develop opportunities for users to interact with control systems. Thus, implementing model predictive control, ANN and supervisory control systems is very useful to achieve such an objective. The paper also highlighted the main KPIs related to occupants' comfort and IEQ quantification. Moreover, this review article has presented some of the significant experimental sensors used to evaluate the indoor environmental conditions in a space. Moreover, these sensors can indicate the pattern of the interaction between occupants and the indoor environment. Furthermore, this paper has shown that much research is done in the area of smart technologies, control systems, BEMS, IEQ and occupant comfort. However, there has been only few research done on building user interfaces advancements, their application in SBs, the way users interact with these interfaces, as well as the resulting comfort and energy performance. Thus, future research is needed to understand how occupants interact with building interfaces and the impact of operations and interfaces on occupants' satisfaction and perceived control. Additionally, there is a need for developing KPIs that quantify the smart technologies in SBs.