Introduction

With the continuous advancement of technology and the implementation of dual carbon goals, in the creation of indoor thermal environments, people are pursuing not only comfort but also low-carbon operations.

In building operation modes, buildings operating in natural ventilation (NV) mode without AC system rely on the thermal and wind pressure generated by the difference in indoor and outdoor air density for ventilation. Occupants can only maintain their comfort to an extent by opening doors and windows, or using fans during summer1. However, with the frequent occurrence of extreme weather2 and the popularization of AC3, NV alone can no longer meet occupants’ needs for a healthy and comfortable indoor environment. Conversely, in AC mode buildings, the continuous operation of AC ensures occupant comfort but leads to considerable energy wastage. In developing countries in Asia, more and more split air conditioners are being used in office and residential buildings4,5. In residential buildings with split air conditioners, occupants choose to turn on the AC when the indoor temperature exceeds the acceptable temperature range. The operation mode switches between AC and free running based on the decisions of the occupants. In this case, the building is considered to operate in mixed-mode (MM)6,7.

In the three operation modes mentioned above, whether the AC is available or adjustable makes a difference in the occupants’ personal control, which may lead to a new level of adaptation to the thermal environment. This falls within the scope of the adaptive thermal comfort theory.

The adaptive thermal comfort theory posits that individuals can actively adapt to the thermal environment at physiological, behavioral, and psychological levels8, and it is increasingly applied in thermal management. When initially defined, personal control was categorized into three types: available control, exercised control and perceived control9. Among these:

  1. (1)

    Available control means the control methods provided by the building, such as air conditioning (AC), fans and windows.

  2. (2)

    Exercised control refers to the actions through which the occupant is actively operating the control. such as using AC, using fans and adjusting windows.

  3. (3)

    Perceived control signifies the occupant’s evaluation of their capacity to manage the thermal environment. The term has roots in fields of personality psychology, and it has received significant backing from neuroscience investigations10. The impact of perceived control on thermal responses has been validated through both field and laboratory experiments11,12. However, the specific temperature conditions that influence thermal responses through perceived control, as well as its impact on the energy consumption of buildings operating in natural ventilation and mixed-mode, remain undetermined.

Firstly, perceived control’s impact on thermal responses has been confirmed under cold or warm conditions, while its effect is not significant in neutral environments. Therefore, there might be a threshold temperature at which perceived control begins to significantly affect thermal responses. For summer (warm) conditions, Zhou et al.13 highlighted the positive impact of personal control in a “warm” thermal environment, Luo et al.14 suggested the effect of perceived control to be more pronounced in serve thermal environments (21 ℃, 31 ℃) as compared to more near-neutral ones (26 ℃, 28 ℃, 24 ℃). Nevertheless, neither of them precisely delineated the threshold temperature at which perceived control begins to influence thermal responses as temperatures rise or deviate from the neutral range.

Secondly, the direct impact of perceived control on cooling energy consumption in both NV and MM buildings has not been quantitatively assessed. Yun12 calculated the cooling energy consumption when the building operating at the neutral temperatures of the occupants with high and low perceived control respectively. She found that increasing occupants’ perceived control level could reduce building cooling energy consumption by 9%. However, there were still numerous NV buildings. To maintain comfort level throughout warm months, these buildings are equipped with electric fans, the operation of which contributes to a certain level of carbon emissions. Meanwhile, MM is increasingly being utilized and advocated15,16 due to its more significant energy-saving advantages over AC mode, as well as the broader range of acceptable temperatures tolerated by their residents compared to what is predicted by the PMV/PPD model17,18,19. Consequently, it is crucial to establish the energy saving capabilities of perceived control in both NV and MM buildings.

Moreover, according to existing research, perceived control can be influenced not only by the available control in the building20, but also by the state of occupants. Regarding the state of occupants, Karjalainen21 and Schweiker22 found that indoor personnel density and room size affected perceived control. Furthermore, psychological research suggests that factors such as age23, cultural background24, and income25 also impact the perception of control. To mitigate these potential disruptors, the student dormitory with relatively fixed personnel density and room size was chosen for research, and the students of similar age, living habits and income levels in dormitory building were selected as subjects.

The main objectives of the survey are to quantify the working temperature range and carbon reduction effect of perceived control in both NV and MM operation, and then provide pertinent guidance for the design and operation of such modes of buildings.

Methods

The methods used in the research include field measurements, questionnaire surveys, statistical analysis, and energy consumption simulation. All methods were carried out in accordance with relevant guidelines and regulations.

Survey ___location and time

Situated in the northwestern quadrant of Henan Province, 35°24’N, 113°21’E, lies Jiaozuo city. A field survey was conducted throughout the duration of summer, spanning from June 1st to August 31st, 2021. On each survey days, the researchers visit student dormitories at random times between 8 AM and 10 PM.

Building and subjects

The dormitory building of a university was selected as the test subject. Students in the dormitory were selected as the research objects. Starting from July 25th, split air conditioners were uniformly installed in the dormitories. Once all installations were complete, the survey continued. From June to August 2021, a total of 61 days of field research were conducted, with 10 to 15 samples obtained every day on the premise of ensuring uniform survey time, dormitory floor and orientation. According to statistics, A total of 922 valid samples were obtained. Bounded by the date of installation of the split air conditioners, the building operation mode before installation of the air conditioner was defined as NV, and after installation was defined as MM. 399 samples belonged to the NV group and the MM group accounted for 523 samples. It is important to note that the same subjects could be present in both groups. The average age of the obtained samples was 21.9. The male to female ratio was 0.8. All subjects resided Jiaozuo City for more than a year, thus were accustomed to the local climate.

Regarding available control, each dormitory consisted of a balcony and two ceiling fans. Students were free to open and close doors and windows. Excluding air conditioning, the dormitory facilitated the same available control in both operation modes. All available controls are freely operated by the occupants.

Field measurements and questionnaire

The study protocol was submitted to and approved by the Ethics Review Committee of Henan Polytechnic University, which confirmed that the study complies with current international and domestic ethical standards. During the field investigation, students in the dormitories were invited to voluntarily participate in the questionnaire survey. Informed consent was obtained from all subjects for both study participation and the publication of identifying information in an online open-access journal.

The process of the field survey is as follows. Before the measurements, the students were thoroughly informed about the purpose, procedures, potential risks, and expected benefit. If they agreed to voluntarily participate, the staff began setting up the testing equipment. After the equipment was stabilized, parameter measurements and the questionnaire survey were conducted simultaneously. The completed questionnaires were immediately submitted once finished, and the environmental parameter testing took about 15 min26.

In order to make the survey results reflect real-life conditions, students’ clothing, ___location and use of control devices were not restricted. Furthermore, participants were assured of their right to withdraw from the test at any stage without any consequences affecting their legal rights or interests.

Outdoor dry bulb temperature (Tout), outdoor wet bulb temperature (Tout’) and relative humidity (RHout) was measured by automatic meteorological station. These measurements were taken per minute on the top floor of the dormitory building.

Indoor thermal environment parameters of the occupants were measured, and each data was measured once following the Class II method outlined in the ASHRAE 55 standard27. The measured indoor environmental parameters included air temperature (Ta), relative humidity (RH), and air velocity (v), all recorded near the occupants at heights corresponding to their posture − 0.6 m for seated and 1.1 m for standing positions. Three local skin temperatures included arm, forehead, and ankle were also measured. The accuracy and valid range information of the instrument accords with the standard of ISO 772628. Local skin temperature was measured by the LO184-JTNT-A, typically completed within 1 to 3 s25.

The questionnaire collected the following information:

  1. (1)

    Background information including gender, age, activity level, and clothing. The clothing and activity status were ticked in the questionnaire.

  2. (2)

    Subjective thermal responses, encompassing the thermal sensation vote (TSV)24 and the thermal acceptability vote (TAV)29.

  3. (3)

    Perceived control over the thermal environment, which was evaluated using a 4-point perceived control vote (PCV) scale. “Complete control” (PCV = 3) represents the perception of total control over the thermal environment. “High degree of control” (PCV = 2) suggests substantial, though not total, control. “Slight degree of control” (PCV = 1) represents the belief in minimal control, with most desires for control remaining unfulfilled. Lastly, “No control” (PCV = 0) signifies a perception of complete inability to influence the thermal environment. This 4-point voting format ensures that the samples are as evenly distributed as possible across each vote, without causing cognitive burden for the participants.

  4. (4)

    AC setting temperature: When the AC was turned on, participants were asked to check the AC setting temperature in the questionnaire. The options were single-choice, ranging from 16 ℃ to 30 ℃.

Data analysis

In the process of computing the PMV, the mean radiant temperature (Tmr) was calculated using air temperature (Ta), black-globe temperature (Tg), and air velocity (v) according to the result of onsite measurement, as shown in Eq. (1)30. The assessments of clothing insulation and metabolic rates adhered to the ASHRAE 5524 and GB50736-2012 Design Standard for Indoor Thermal Environment Control in Buildings31, determined by occupants’ ticked clothing and activity status in the questionnaire. Sedentary occupants, those engaged in light activity while seated, those performing mild activities while standing, and those moderately active in a standing pose were assigned metabolic rates of 1 met, 1.2 met, 1.4 met, and 2 met, respectively24.

$$T{}_{{{\text{mr}}}}={\left[ {{{\left( {{T_{\text{g}}}+273} \right)}^4}+2.5 \times {{10}^8} \times {v^{0.6}}\left( {{T_{\text{g}}} - {T_{\text{a}}}} \right)} \right]^{_{{^{{1/4}}}}}} - 273$$
(1)

In the interest of data accuracy and reducing the influence of extreme voting, binning and weighting methods were utilized to scrutinize the data. The indoor air temperatures were binned into 0.5 ℃ increments to calculate the average value of thermal response votes. Followed by this, the resultant regression lines underwent covariance analysis testing for differences. All statistical analyses were rendered through SPSS v 26.0. The p-value was used to describe whether the data were significantly correlated or different. A p-value of less than 0.05 would imply that the two comparative groups were significantly different.

Simulation methods

To accurately illustrate the carbon reduction impact of perceived control, the surveyed building was modeled in DesignBuilder, and its cooling energy consumption was simulated in EnergyPlus. Initially, the actual parameters of the building were applied in DesignBuilder modeling to ensure the precision of the indoor thermal environment simulation. Following this, relying on meteorological data collected during the field survey and thermal responses from the questionnaire, the AC operating schedule and setting temperatures for occupants with varying levels of perceived control were computed. Subsequently, the corresponding energy consumption was simulated.

Modeling with designbuilder

In the software DesignBuilder, the building was modeled according to the ___location and structure information of the surveyed building. The building was oriented towards the south, skewed by 7° towards the east. The structure was seven stories high with each floor measuring 3.0 m. The total construction area was 2198.4 m2. The shape coefficient of the building was about 0.17, and the window wall ratio was about 0.55. Key details such as the indoor conditions and the building’s exterior as derived from the DesignBuilder modeling are depicted in Fig. 1.

Fig. 1
figure 1

The internal and external conditions of the dormitory building.

Full size image

The building maintenance structure complies with GB50189-2015 Design Standard for Energy Efficiency of Public Buildings32, and the structural parameters are shown in Table 1. Those construction information was entered into the model for indoor thermal environment simulation. The first floor was dedicated to off-campus business, while the remaining six floors functioned as dormitories. Notably, detailed modelling and energy consumption simulation were carried out only for the dormitory section.

Table 1 Thermal properties of Building components.
Full size table

Simulating with energyplus

Developed by the Department of Energy, EnergyPlus is grounded on the principle of thermal equilibrium and has received acknowledgment for its accuracy in predicting energy consumption within buildings33,34.

To reflect real-world conditions in the simulation, the cooling capacity (3,510 W) and Coefficient of Performance (COP) (4.68) for the AC used in the surveyed building were input to EnergyPlus. After entering the cooling capacity and COP, the electrical energy required to run the AC at 26.0 ℃for 24 h was simulated and compared with actual conditions, completing the model verification.

The AC operation scheme was obtained by simulation and calculation of the data obtained from the field investigation. The outdoor climate parameters and the thermal response of occupants were collected during the field investigation. According to the data in the field investigation, the upper limit of the acceptable temperature of the occupants can be calculated, that is, the starting temperature of the AC. The differences in AC usage duration among occupants with varying levels of perceived control can be determined by comparing the temperatures at which they need to turn on the AC with the room temperatures under natural ventilation conditions.

Results

Outdoor thermal environments

Despite its classification as a cold climate area according to GB/T 50,785 China’s Thermal Design Code for Civil Buildings23, the Jiaozuo City recorded a daily mean temperature of 28.5 ℃ during the field survey period, fueling a heavy demand for HVAC equipment among local inhabitants. On July 25, 2021, the surveyed school installed split air conditioners for the students’ dormitories. After the installation, occupants could decide the AC’s start, stop, running period, and setting temperature according to their needs. This operating mode was considered MM. Before the installation, the operating mode was defined as the NV mode.

The field study was conducted for 35 days when the dormitory building operated in NV mode (n = 399), the daily mean outdoor temperature was 28.8 ℃. The field study was conducted for 36 days when the dormitory building operated in MM (n = 523), the daily mean outdoor temperature was 28.5 ℃. According to the independent-sample t test results, there was no statistically significant difference (P = 0.656) in the daily mean outdoor temperature during the two modes of operation. Therefore, occupants had comparable outdoor thermal experiences in both operation modes.

Indoor thermal environments

The results of independent-sample t tests show no significant differences in air temperature and relative humidity (P < 0.001) across the two operation modes. Under NV operation, the average air temperature and relative humidity were 32.0 ℃ and 53.8% respectively. Under MM operation, the average temperature was at 27.8 ℃ and relative humidity at 61.7%. The difference in air velocity was found to be non-significant (P = 0.332). During operation in both modes, the average air velocity was maintained at 0.6 m/s. The variations in both air temperature and humidity were predominantly driven by the use of AC, and to some extent, affected by the management of building elements such as doors, windows, and fans.

Local skin temperature

All participants in the study had resided in Jiaozuo City for over a year, thus having adapted to the local climate. However, varying accessibility to AC resulted in a divergence in indoor air temperatures between NV and MM operation periods. However, previous studies have indicated that exposure to disparate thermal environments can affect physiological adaptations, primarily seen in thermoregulatory capacity. These thermoregulatory results can be exhibited in local skin temperature25. To better understand whether the installation of AC generated a new level of physiological adaptation, samples from the initial ten days post-AC installation and the final ten days prior to survey termination were selected for comparison. The standard effective temperature (SET) considers the occupants’ activity and clothing, the influence of behavioral adaptation can be eliminated to a certain extent. Thus, SET was chosen as the benchmark for temperature division. Analyzing local skin temperature within the same SET range aids in discerning any potential changes in the level of physiological adaptation represented by thermoregulation levels.

The SET was computed using all measurable physical quantities and personal factors with the two-node model35. Statistical analysis and independent-sample t tests were conducted for the three points’ local skin temperature at every 2 ℃ SET segment of the two groups of occupants. The minimum, maximum, quartile, mean, and median local skin temperature for each group, as well as the t test results, are shown in Fig. 2, where * indicates a significant difference between the two groups of data (P < 0.05), ** indicates a very significant difference (P < 0.01), and ns indicates no difference between the two groups of data (P > 0.05). As can be seen, there was no difference in local skin temperature within the same SET segments during this period after AC was installed. Hence, AC installation did not impact the level of occupants’ physiological adaptability.

Fig. 2
figure 2

Difference of local skin temperature in every 2 ℃ SET interval.

Full size image

Thermal responses

During the survey, a 7-point thermal sensation vote (TSV) (ranging from − 3 for cold, -2 for cool, -1 for slightly cool, 0 for neutral, 1 for slightly warm, 2 for warm, to 3 for hot) and a 4-point thermal acceptability vote (TAV) (ranging from − 2 for totally unacceptable, -1 for barely unacceptable, 1 for barely acceptable, to 2 for totally acceptable) were collected through questionnaires. Results of independent-sample t test indicates significant differences in both TSV and TAV among occupants in the two operation modes (P < 0.001, P < 0.001). In NV mode, a mere 10.9% of occupants reported “moderate” thermal sensations, compared to 56.8% in MM. In terms of thermal acceptability, 52.4% of MM occupants deemed the thermal environment totally acceptable, compared to only 33.3% in NV mode.

During MM operation, occupants could utilize AC, leading to a divergence in the indoor air temperature compared to NV mode. Thus, a comparison of the thermal responses of occupants in the two operation modes within the same air temperature (Ta) range can better elucidate the variations in their thermal adaptation levels. Both the TSV and TAV were statistically analyzed and tested for every 2 ℃ segment in air temperature. Both the minimum, maximum, quartile, mean, and median TSV for each group and the t test results are shown in Fig. 3, where ** indicates a very significant difference (P < 0.01) of the two groups of data, and ns indicates no difference (P > 0.05).

Fig. 3
figure 3

Results of statistical and independent-sample T test for TSV and TAV.

Full size image

As depicted, there were no statistically significant differences in TSV and TAV between NV and MM groups within the 26–28 ℃ and 28–30 ℃ temperature ranges. However, significant differences in TSV and TAV emerged between the groups in the temperature range of 30–32 ℃. Within this range, the average TSV of MM group (0.8) was lower than of NV group (1.7), and the average TAV of MM group (1.4) was higher than of NV group (1.0). To interpret, occupants in mixed-mode reported more moderate thermal sensations and demonstrated higher acceptability than in natural ventilation mode in 30–32 ℃ environment. As confirmed in section “Local skin temperature”, AC installation did not alter the physiological adaptation level, suggesting the observed mitigation in thermal response can primarily be attributed to an improvement in occupant’s behavioral and psychological adaptation, stemming from ‘perceived control.’

Perceived control

During the survey, a 4-point perceived control vote (PCV) was collected through questionnaires, ranging from 0 for no control, 1 for slight degree of controllable, 2 for high degree of control, to 3 for complete control. A statistical analysis of the PCV among occupants in both operation modes was carried out. Occupants in MM reported an average PCV of 2.1, higher than the 1.3 average among NV mode occupants. According to the result of independent-sample t test, this disparity in PCV was found to be significant (P < 0.001), indicating that occupants felt a greater sense of control over the thermal environment in MM compared to NV mode.

Effect of perceived control

To identify the impact of perceived control on the thermal response in the two operation modes, a correlation test was conducted between the PCV and both TSV and TAV. The resultant p-values reveals that in NV mode, a significant negative correlation existed between PCV and TSV (P < 0.001). Conversely, in MM, PCV exhibited a significant positive correlation with TAV (P = 0.003).

To further investigate the impact of the level of perceived control on thermal response and to meticulously analyze thermal response at identical temperatures, weighted regression analyses were performed separately for the TSV of the NV samples and the proportion of thermally acceptable (TA) of the MM samples in relation to air temperature.

Firstly, the analysis results for the NV samples are presented. The samples in both operation modes were categorized based on their PCV. A linear regression analysis was subsequently carried out between the TSV and air temperature (Ta) for each group. The analysis outcomes are depicted in Fig. 4 (A), Eqs. (2), (3), (4), (5).

$$No:TSV = 0.222T_{a} - 4.865\left( {R^{2} = 0.485,P = 0.000} \right)$$
(2)
$$Slight\,\deg ree:TSV = 0.237T_{a} - 5.954\left( {R^{2} = 0.661,P = 0.000} \right)$$
(3)
$$High\,\deg ree:TSV = 0.208T_{a} - 5.319\left( {R^{2} = 0.687,P = 0.000} \right)$$
(4)
$$Complete:TSV = 0.442T_{a} - 12.073\left( {R^{2} = 0.774,P = 0.000} \right)$$
(5)

As observable from Fig. 4 (A), for NV mode, the ‘no control’ group predominantly fell within the warm environment (Ta > 30.0 ℃). Given the uniform control conditions over the thermal environment for NV occupants, the grouping was based on a subjective assessment of perceived control, which suggests that occupants were more prone to feeling out of control in warm conditions. The linear regression coefficient reveals the extent of occupants’ thermal sensitivity to temperature fluctuations17. Covariance tests are applicable for evaluating the statistical significance of differences between two linear regression coefficients representing thermal sensitivity. According to the results of the covariance test between the pairwise of Eqs. (2), (3), (4), (5), no significant difference was present in the regression coefficient (P > 0.05), approximately 0.2, across three regression groups: no control, slight degree of control, and high degree of control. This indicates that occupants in NV mode had similar thermal sensitivity across these three votes. With an increase in perceived control level from slight to high degree, and from high degree to complete, the TSV decreased by about 0.6 and 0.3 vote units, respectively. Notably, the regression coefficient for the complete control group (approx. 0.4) was significantly larger than that of the other three groups (P < 0.001), suggesting increased sensitivity to temperature changes among these occupants. Furthermore, in conditions where TSV = 0, occupants were considered to have attained a thermoneutral state, i.e., feeling neither cold nor hot17. The calculated neutral temperatures for occupants under natural ventilation mode demonstrated an increment parallel to the rise in perceived control level, registering temperatures at 21.9 ℃, 25.1 ℃, 25.6 ℃, and 27.3 ℃.

Subsequently, the analysis results for the MM group are presented. The samples were grouped with their PCV, and the TA was calculated for each group. Both “barely acceptable” and “totally acceptable” votes were counted as TA. To explore the effect of perceived control on the occupants’ acceptable temperature range, the MM samples were grouped according to PCV. The TA were analyzed within each 0.5 ℃ increments of air temperature. With this, the binomial regression was conducted using the corresponding air temperature (Ta), as demonstrated in Fig. 4 (B) and Eqs. (6), (7), (8). The regression was not completed for the ‘no control’ group because the sample size for this group was too small (n = 7).

Fig. 4
figure 4

Relationship between TSV, TA and Ta under each PCV. (A) Relationship between TSV and Ta under each PCV for NV group. (B) Relationship between TA and Ta under the four PCV for MM group.

Full size image
$$Slight:TA = - 0.236T_{a}^{2} + 6.784T_{a} + 60.362\left( {R^{2} = 0.657,P = 0.000} \right)$$
(6)
$$High:TA = - 1.802T_{a}^{2} - 95.4T_{a} - 1175.626\left( {R^{2} = 0.254,P = 0.000} \right)$$
(7)
$$Complete:TA = - 1.673T_{a}^{2} + 88.818T_{a} - 1086.517\left( {R^{2} = 0.132,P = 0.000} \right)$$
(8)

An air temperature deemed acceptable by 80% of the occupants was considered acceptable26. It was calculated that air temperatures below 25.4 ℃ were deemed acceptable by the group with a slight degree of control. For the group with a high degree of control, the acceptable temperature range was 24.5 ℃ to 28.4 ℃. For the complete control group, the acceptable temperature range was 23.8 ℃ to 29.3 ℃. Consequently, the upper limit of occupants’ acceptable temperature range demonstrated an increment parallel to the rise in perceived control level in MM.

Discussion

The operating temperature range of perceived control

Laboratory studies have consistently shown that perceived control can positively influence occupants’ thermal sensations in severe thermal environments. Zhou et al.14 examined the impact of perceived control on thermal responses under three scenarios: the absence of AC, freely usable AC, and charged AC. The study concluded that AC control notably ameliorates thermal sensations in relatively warmer environments. In a different study, Luo et al.15 conducted experiments at various temperatures (31 ℃, 28 ℃, 26 ℃, 24 ℃, 21 ℃). They deduced that perceived control significantly influences thermal sensations in both warm and cool conditions (21 ℃ and 31℃). However, the specific operational temperature range for perceived control still requires clear identification.

The PMV model, typically applied to steady-state and neutral thermal environments, does not account for the role of “adaptation36”. Comparing TSV and PMV of the occupants with perceived control (PCV = 1, 2, 3) at the same air temperature can determine whether thermal adaptation caused by perceived control occurred. All samples with perceived control in NV mode were grouped. According to Fanger’s equation, the PMV value was computed through MATLAB programming. The linear regression analysis of TSV and PMV with air temperature (Ta) was performed, respectively. The analysis results are shown in Fig. 5 (A), Eqs. (9), (10).

$$PMV = 0.398T_{a} - 10.954\left( {R^{2} = 0.967,P = 0.000} \right)$$
(9)
$$TSV = 0.261T_{a} - 6.162\left( {R^{2} = 0.785,P = 0.000} \right)$$
(10)
Fig. 5
figure 5

Determination of the operative temperature range for the NV mode. (A) Relationship between TSV and Ta in NV mode. (B) Difference between TSV and PMV in every 2 ℃ in NV mode.

Full size image

Figure 5A demonstrates that TSV consistently aligns more closely with neutral (0) than PMV in the warm conditions, indicating that perceived control positively affected occupants’ thermal sensations. Equations (9), (10) intersected at an air temperature of 30.5 ℃, which can be interpreted as the critical temperature point. Subsequently, the independent-sample t test was performed for TSV and PMV of the sample with perceived control (PCV = 1, 2, 3) in each 2 ℃ temperature segments, as depicted in Fig. 5B. It emerges that when the indoor air temperature was lower than 30.5 ℃, the difference between TSV and PMV was not statistically significant (P = 0.819, 0.410). Nevertheless, When the indoor air temperature exceeded 30.5 ℃, PMV and TSV began to show a significant difference (P < 0.05). Occupants’ actual thermal sensation in a warm environment was lower than that predicted by the PMV model. The deviation between TSV and PMV becomes more evident as the temperature increases, implying that 30.5 ℃ may be considered the threshold temperature beyond which perceived control began to impact thermal sensation positively in NV mode.

A similar analysis was used to identify the temperature range where perceived control positively affected thermal sensation in MM. All samples with perceived control (PCV = 1, 2, 3) in MM were grouped. The linear regression analysis of TSV and PMV with air temperature (Ta) was performed, respectively. The analysis results are shown in Eqs. (11), (12).

$$PMV = 0.326T_{a} - 8.934\left( {R^{2} = 0.843,P = 0.000} \right)$$
(11)
$$TSV = 0.115T_{a} - 2.825\left( {R^{2} = 0.702,P = 0.000} \right)$$
(12)

Equations (11), (12) intersected at 29.0 ℃. Furthermore, the independent-sample t test results demonstrate a significant difference between TSV and PMV at every 2 ℃ segment (P = 0.000, 0.000, 0.003, 0.004, 0.044), TSV was always closer to neutral (0) than PMV. As such, it’s deduced that perceived control contributed positively to thermal sensation across the entire temperature range (23.3 ℃ ~ 32.7 ℃) utilized in the investigation in MM.

In conclusion, perceived control has not only enabled NV mode occupants to better adapt to warm environments exceeding 30.5 ℃, surpassing the predictions of the PMV model, but it has also brought the thermal sensation of MM occupants closer to neutral across the entire surveyed temperature spectrum, ranging from 23.3 ℃ to 32.7 ℃.

Carbon reduction potential of perceived control

In both operation modes, enhancing the level of occupants’ perceived control can generate significant energy-saving benefits and contribute to carbon reduction.

In natural ventilation mode

Under NV conditions, the use of fans by occupants varied according to their perceived level of control—92.1% for the no control group, 90.1% for the slight degree of control group, 89.0% for the high degree of control group, and 75.0% for the complete control group. There was a notable decrease in fan usage—17.1%—as perceived control progressed from none to complete. This statistical data suggests that, in NV mode, bolstering the perceived control level of the occupants can curtail their reliance on fans. This, in turn, reduces overall electricity consumption and carbon emissions in building operations to a notable degree.

In mixed-mode

In MM, occupants had the autonomy to switch the AC on and off and to adjust its setting temperature and operating hours to meet their needs. Xue et al.37 conducted research on residential buildings in China, revealing that occupants in MM buildings activated AC only when temperatures exceeded the upper limit of acceptable temperature. Similarly, Loengbudnark38 also found that delegating the task of turning on the equipment to the occupants could effectively reduce the energy consumption of the operation of the automated building. Considering the AC operating characteristics of MM occupants, combined with the results of questionnaire survey, the building cooling energy consumption required by each PCV occupants was simulated. In the simulation, the AC operating schedule and setting temperature are the necessary input variables.

Firstly, to determine the AC operating schedule, the indoor air temperature of the building under NV operation during the selected two days was simulated and compared with the upper limit of acceptable temperature range for each PCV group in the MM mode. The duration that the simulated air temperatures surpassed upper limit of acceptable temperature range dictated the AC operation time for each group.

Due to the variability of outdoor climate conditions, we chose typical days for energy consumption simulation that closely matched the average daily outdoor temperature during the entire survey period. Additionally, considering the noticeable indoor occupancy variations between weekdays and weekends, consecutive weekdays and weekends were selected for the simulation. During the MM operation period, the daily mean outdoor temperature was 28.0 ℃. Accordingly, August 13th (Friday) and August 14th (Saturday), both with an average daily outdoor temperature of 28.0 ℃, were selected for the simulation. The outdoor climate parameters, recorded hourly during the field survey from August 13 to August 14, 2021, and depicted in Fig. 6, were entered into EnergyPlus.

The room occupancy rates were then established in accordance with the institution’s summer work schedule. The working hours were from 8 AM to 12AM and from 3 PM to 6 PM on weekdays, reducing occupancy in dormitories. Each dormitory housed six students, and based on their typical habits, the personnel schedule was configured in EnergyPlus. This configuration reflected 0% occupancy during working hours and 100% during the rest. On weekends, the occupancy was varied, set to be 50% from 8 AM to 10 PM and 100% for the remaining hours.

Fig. 6
figure 6

Outdoor environmental parameters during the simulation period.

Full size image

To simulate the indoor temperature under NV conditions over these two days, it was necessary to input the ventilation rate of the room into the model. According to the GB 50,736 − 2012 Design Standard for Indoor Thermal Environment Control in Buildings28 and the GB 50,099 − 2011 Code for Design of School39, the ventilation frequency for dormitories should be more than 1 ACH. Based on empirical studies of university dormitories in China, the ventilation frequency under NV conditions can reach 3.74 ACH40, meeting the national standards. Therefore, a ventilation frequency of 3.74 ACH was input into the model to simulate the indoor temperature under NV conditions. Figure 7 illustrates the indoor temperatures simulated for the building under outdoor climatic conditions on August 13th and 14th, operating in NV mode.

Fig. 7
figure 7

Indoor temperature in a naturally ventilated state.

Full size image

According to Section “Effect of perceived control”, in MM, the upper limits of acceptable temperature range were 25.4 ℃ for the slight control group, 28.4 ℃ for the high control group, and 29.3 ℃ for the complete control group. The simulation calculated the AC usage schedule based on these thresholds. According to calculations, removing the class time of weekdays, occupants with a slight degree of control required 17 h of AC operation on weekdays, compared to 9 h for those perceiving high control and 8 h for complete control. Weekend operation times increased to 24, 16, and 13 h for the respective groups.

To determine the AC setting temperature of each PCV sample for simulation, the MM samples were grouped according to PCV, and their response to the question “What is your recent air conditioning setting temperature?” was analyzed. The result of independent-sample t test results highlighted no statistically significant difference in the AC setting temperature among occupants under the three PCVs (P > 0.05). The average AC setting temperature for occupants across all PCVs was 26.0 ℃, in compliance with the requirements of the national standard28. Thus, in the simulation, the AC was all set to operate at 26.0 ℃ when occupants needed it.

Subsequently, the cooling energy consumption attributable to occupants under each PCV was simulated. An emission factor, which delineates the relationship between energy consumption and carbon emissions, was employed to compute carbon emissions resulting from energy usage during building operations41. Carbon emissions at the building operation equate to the energy consumption multiplied by their corresponding emission factors42. In 2021, the carbon emission factor for electricity in China was 0.5703kgCO2/kWh43. The simulated cooling electricity consumption and the calculated carbon emissions are documented in Table 2.

Table 2 Cooling electricity consumption and carbon emissions under different PCV.
Full size table

Taking into account the psychological and behavioral adaptability facilitated by perceived control, MM occupants attaining a perception of complete control can decrease AC carbon emissions by 46.6% on weekdays and 37.7% on weekends, compared to those with a slight degree of control. Enhancing occupants’ perceived control can significantly minimize AC operating duration, subsequently yielding a notable reduction in operational carbon emissions. Compared to the simulation results from Yun’s12 study in an AC building, perceived control can yield more significant energy-saving and carbon reduction effects in mixed-mode buildings. Hence, perceived control has been demonstrated to hold greater potential for carbon reduction in developing countries with a low prevalence of air conditioning systems and extensive use of split air conditioners.

Conclusion

In light of present studies, this study aimed to provide recommendations and guidance for building design and user energy habits where natural ventilation and split air conditioners are widely used. Upon confirming the positive influence of perceived control on thermal sensations, the temperature range of its impact was accurately quantified. Furthermore, the effects of perceived control on carbon reduction in both natural ventilation and mixed-mode buildings were quantified by statistics and simulation. The main conclusions are as follows:

  1. (1)

    Despite similar daily mean outdoor temperatures as experienced in natural ventilation mode and mixed-mode, lower indoor air temperatures and higher relative humidity were recorded in mixed-mode operation.

  2. (2)

    Despite the same outdoor thermal experience and physiological adaptation level, occupants in the mixed-mode had more moderate thermal sensations and higher thermal acceptability in the warm environment (30 ℃ ~ 32 ℃) than in natural ventilation mode due to their higher perceived control level.

  3. (3)

    Occupants perceived a higher level of control in mixed-mode than in natural ventilation mode. Perceived control influenced occupants’ thermal sensation in natural ventilation mode and thermal acceptability in mixed-mode.

  4. (4)

    In natural ventilation mode, an increased level of perceived control allowed occupants to achieve a thermoneutral state at higher temperatures (21.9 ℃, 25.1 ℃, 25.6 ℃, 27.3 ℃). In mixed-mode, improved perceived control led to an increased upper limit of acceptable temperature by 3.9 ℃ at most.

  5. (5)

    The specific temperature ranges in which perceived control positively impacted thermal sensation varied between the two operation modes. In natural ventilation mode, perceived control only had an effect in warm environments (> 30.5 ℃), but showed no notable impact in neutral environments. In mixed-mode, perceived control positively influenced thermal sensation across the entire temperature range encompassed by the field study (23.3 ℃ to 32.7 ℃).

  6. (6)

    Enhancing occupants’ perceived control level can reduce carbon emissions in both operation modes. By improving perceived control in natural ventilation mode, fan usage was reduced by 17.1%. In mixed-mode, energy savings peaked at a rate of 46.6%.

Limitations and future research

In an attempt to isolate key variables and scrutinize the perceived control’s mechanisms on thermal response, this research focused on a demographic of college students. Notwithstanding, the thermal perception and perceived control among occupants is impacted by factors such as age, income, culture, and education level, thereby limiting the universal applicability of our conclusions. Future research should delve into perceived control’s role on the thermal responses among varied demographic categories marked by different ages, incomes, and cultural and educational backgrounds. This approach bolsters efforts to amplify the psychological adaptation’s carbon reduction effect across broader application scenarios. Furthermore, empirical research in an expansive variety of building types would provide robust evidence to corroborate the mechanisms of perceived control further.