Thermal resilience to overheating assessment in a Belgian educational building with passive cooling strategies during heatwaves and power outages

. Airtight and highly insulated educational buildings are subjected to overheating risks, even in moderate climates, due to unforeseeable events like frequent heatwaves (HWs) and power outages (POs) leading to heat-stress and negative impact on the health conditions and cognitive performance of the students. The focus of this paper is to evaluate thermal resilience for two lecture rooms equipped with the low-energy cooling strategies natural night ventilation (NNV) and indirect evaporative cooling (IEC). To assess the thermal resilience to overheating, the lecture rooms were tested with and without passive cooling strategies for 3 Typical meteorological years (TMYs), 3 severe HWs and those 3 HWs + POs. Results evaluating the existing indicators unmet degree hours, indoor overheating degree (IOD), ambient warmness degree (AWD), and overheating escalation factor (α IOD ) demonstrated that with passive cooling strategies the two test lecture rooms have good thermal resilience during TMY and HW periods (except long-term severe HW), with 18% higher unmet degree hours during HWs. Lecture room with heavier thermal mass demonstrated higher thermal resilience to overheating in long-term assessment. Furthermore the need to develop a holistic resilience indicator taking into account building and system parameters was also pointed out in this study.


Introduction
In June and July of 2019, Europe experienced historical records for the highest temperatures, and heatwaves (HWs) especially in the western-central regions [1], with excess deaths associated with heat-stress during these HWs. Severe HWs like 2019 are projected to become the norm by 2050, which could lead to a 257% increase in heat-related mortality in vulnerable communities [2]. Thus, warmer summers and frequent HWs will increase the cooling energy need as well as the overheating risk in buildings [3]. Additionally, high cooling energy demand during HWs puts extreme pressure on the electric grid and thus, HWs are often accompanied by power outages further exacerbating overheating risks [4]. Heating, Ventilation and Air conditioning (HVAC) systems installed in buildings to provide occupants with comfortable indoor conditions are energy intensive [5]. Globally, and especially in the EU, a shift towards designing energy efficient buildings was put in motion through revision of building codes and standards [6] to meet the thermal comfort needs of occupants at reduced energy costs [7]. Thus, current design methods focus on energy efficient solutions. Energy efficient building envelope parameters include high insulation, airtight envelopes [8], improved glazing [9], and solar shading [10] whereas passive cooling * Corresponding author: abantika.sengupta@kuleuven.be strategies include ground source cooling [11], natural night ventilation (NNV) [12] and indirect evaporative cooling (IEC) techniques [13]. While these buildings and their systems succeed in providing energy-efficient comfortable environments, the airtight and highly insulated envelopes can also retain the heat inside the space creating loads that cannot be completely removed by the passive cooling strategies and thus delaying recovery [14]. The extent to which these buildings and their systems can withstand the thermal shock events, and recover back to the original designed performance, is termed as "thermal resilience" [15]. Therefore, apart from being energy efficient, building performance should also account for thermal resilience. In the framework of IEA EBC's Annex 80 [16], Sengupta et al. [17] demonstrated the thermal resilience of a Belgian dwelling equipped with NNV. The study shows the dwelling recovers 90% faster during HW compared to dwelling without NNV and the peak indoor temperatures decreased by 4.3℃. In another study, Ji et al. [18] found that simultaneously increasing window opening area size, replacing blinds with external shading, applying green roofs and NNV, can improve the buildings' thermal resilience.
The aim of this work is to assess the 'thermal resilience' to overheating of an air-tight, well insulated educational building in Belgium, equipped with IEC, NNV and automated shading. To evaluate the impact of passive cooling strategies on the thermal resilience to overheating, the building was tested with and without the passive cooling strategies. Thermal resilience of the building was assessed against extreme HWs, and HWs in combinations with power outages using existing indicators (Section 3.2). To achieve these objectives, a model of the building was developed in Modelica [19], and building energy simulations (BES) were conducted. This study will give insight on the level of preparedness of today's energy efficient buildings in facing extreme events.

Building and systems' description
The case study building is an educational nZEB at the Technology Campus Ghent, KU Leuven. The building consists of four zones: two lecture rooms (E120-first floor and E220-second floor), a staircase and a technical room. The lecture rooms are identical in design (floor area = 140 m 2 and the volume = 380 m 3 ). E120 has external insulation with a brick external wall, whereas E220 has a lightweight timber frame external wall with the same U-values (Table.1). Thus, E220 has a medium and E120 has a heavy thermal mass according to the EN ISO 13790 [20].  0.65 Figure 1 shows the two lecture rooms and the AHU. The air tightness n50 value of E120 and E220 are 0.41 h -1 and 0.29 h -1 respectively. There are six triple glazed windows on the south-west facade and 4 windows in the North-East façade. The window-to-wall ratio is 26.5% on both facades. The windows on the South-West façade are equipped with internal and external solar shading. South-West windows are equipped with external shading (movable screens), which are controlled automatically (shading is ON when the radiation on the windows is above 250 W/m 2 ). The net energy demand for heating the lecture rooms are calculated in [21]. The annual net demand is 11 kWh/(m 2 .a), achieving the requirements of Passive House standard in school buildings [22]. The building is equipped with an all-air system (100% outdoor air) with balanced mechanical ventilation with a total supply airflow of 4400 m 3 /h. Four Variable Air Volume (VAV) boxes control the airflow of the demand-controlled ventilation system. The airflows are based on CO2 concentration and temperature in the rooms. The two lecture rooms are passively cooled by -(a) IEC at the AHU that cools the supply air by controlling the modular bypass and (2) NNV. For the IEC operation-both the modular bypass and the IEC are part of the Air Handling Unit (AHU) and regulates the air flow either through the IEC or through the modular bypass depending on the external conditions. When the IEC is operation, the AHU supplies the maximum flow rate of 4400 m 3 /h. The maximum capacity of the IEC is 13.1 kW. When the room air temperature exceeds the cooling setpoint by +4℃, IEC is activated. IEC is deactivated when the room temperature reaches setpoint -0.5℃. [23]. NNV relies on the 10 motorized windows with chain actuators (6 on the South-west facade and 4 in the Northeast façade), located 1 m height from the bottom of the floor height. The design of the ventilative cooling system is described in [21]. The total effective operable area of these windows is 4% of the floor area. The control strategy of the natural night ventilation is based on the internal temperature and relative humidity, and external conditions like the outdoor temperature, wind velocity, precipitation. Once open, the window will remain open for at least 15 min. The windows are open between 10 pm to 6 am from 1 st April to 31 st October if the following conditions are met: • Room temperature exceeds both the heating set point (=22°C) and the external temperature +2°C • Maximum room temperature of the previous day exceeds 23°C • External temperature is higher than 12°C • Internal relative humidity is smaller than 70% • There is no rainfall and the wind velocity on site is smaller than 10 m/s Fig. 2. Occupancy of E120 and E220 for a complete week of classes Occupancy for both E120 and E220 during a complete week of classes measured by the building management system (BMS) is shown in Figure. 2. The lecture rooms are in use from Monday to Friday between 8:15 a.m. to 6:00 p.m. with a maximum occupancy of 80 persons or 1.78 m 2 /pers.

Building energy simulations
A Modelica model of the two lecture rooms were developed to conduct building energy simulations (BES) assess the thermal resilience to overheating. Prior to thermal resilience assessment, the simulation model needs to be validated. Long term measurements of temperature, CO2, relative humidity, air flow rates and operation of the fans, VAVs, IEC, solar shading and NNV are conducted for both the lecture rooms [23]. The model was validated experimentally [24]. There are two conditions that the results of the simulation model must meet to be considered validated-(a) The MAE (Mean absolute value of error) < 1°C and (b)RMSE (Root mean squared error) <1.5°C [25]. Comparison of the measured and simulated indoor operative temperatures for lecture room E120 demonstrates MAE of 0.75℃ and RMSE of 0.95°C. This is a good agreement between the monitored and simulated data.

Fig. 4. Overview of simulation scenarios
To evaluate the impact of the IEC, NNV and solar shading on the thermal resilience to overheating, two building operation conditions were tested-(a) building with mechanical ventilation without operation of IEC, NNV and solar shading (b) building with mechanical ventilation and operation of IEC, NNV and solar shading. These two building operation conditions were tested against three scenarios-(a) 3 future TMYs (b) 3 severe HWs and (c) 3 severe HWs accompanied by power outages. 3 power outage durations -(i) 27 days during historical severe HW (ii) 14 days during midterm severe HW and (iii) 45 days during long-term severe HW was simulated. The IEA Annex 80 simulation guideline [26] were adopted to determine the duration of the power outage and the internal (occupancy and equipment) loads during power outage. Figure. 4 illustrates the simulation scenarios.

Weather data
Two types of weather data sets (TMYs and HWs) for 3 periods-(a) historical (2010s based on 2001-2020 data), (b) mid-term (2050s based on 2041-2060 data) and (c) long-term (2100s based on 2081-2100 data) were developed within IEA Annex 80 weather data task force [27]. Average monthly temperature and global solar radiation for the three TMY period has been illustrated in Fig.5. Global solar radiation is reduced in the future TMYs. According to Cutforth et al. [28] this decrease in future solar radiation this can be the consequence of two factors: (a) higher reflectance of solar radiation from increasing aerosol concentrations and sometimes increasing cloudiness, and (b) increase in annual number of precipitation/rain events. Several HWs were detected for each time period. For this study, the 3 most severe HWs from each time period were tested. Table 2 gives an overview of the HWs during which the thermal resilience to overheating was evaluated.

Resilience assessment methods
To assess the thermal resilience of the two lecture rooms, the following existing indicators were tested: (a) Unmet degree.hour [Dh] [29]: This metric is comparable to that of temperature-weighted exceedance hours, a metric defined in ASHRAE Standard 55-2017 [30]. degree.hour metric weighs each hour when the temperature of a zone exceeds a certain threshold by the number of degrees by which it surpasses that threshold. In this study, degree.hours are calculated for a fixed temperature limit (FTL) of 26 ℃ (maximum operative temperature during cooling season in Category II building) [31] and adaptive temperature limit (ATL) calculated [31] given by upper and lower limit of: Ted-i is the daily mean outdoor air temperature for i-th previous day [°C] . ℎ is calculated as follows: Additionally, Method A as described in Annex F of the EN 15251 [32] is selected for the evaluation of summer comfort. For this study, the number or % of occupied hours when the operative temperature is above 26℃, is evaluated. 5% is considered acceptable and 3% is considered good.
(b) Overheating escalation factor (αIOD) [-] [33]: This metric is used to estimate the sensitivity of the building to overheating. It represents the variation in the indoor temperatures when they exceed a chosen thermal comfort temperature limit in a given time period (IOD) as a consequence of the severity of outdoor warmness. The severity of outdoor warmness is quantified using the metric called Ambient warmness degree (AWD18℃). The Overheating escalation factor is defined as: Assuming that the relationship between IOD and AWD18℃ is suitably representable with a linear regression model, the αIOD is the slope coefficient of the regression line. An Overheating escalation factor greater than the unit (αIOD > 1) means that indoor thermal conditions get worse when compared to outdoor thermal stress. On the contrary, an Overheating escalation factor lower than the unit (αIOD < 1) means that a building can suppress some of the outdoor thermal stress. Indoor overheating degree (IOD) [℃] [33]: This multi-zonal indicator is the summation of positive values of the difference between zonal indoor operative temperature Tin,o,z and the zonal thermal comfort limit Tcomf, z averaged over the sum of total number of zonal occupied hours Nocc(z) Where t is the time step [s], i is occupied hour counter [-], z is building zone counter [-], Z is total building zones [-]. Both fixed and adaptive temperature limits can be assumed as comfort thresholds.
Ambient warmness degree (AWD) [℃] [33]: This metric is used to quantify the severity of the outdoor thermal conditions. AWD is the summation of the positive difference between the outdoor air temperature and a base temperature. The selection of the base temperature is context-specific based on the building typology and climate.
where Tout is the outdoor dry-bulb air temperature, Tb is base temperature, N is the number of occupied hours for which Tout > Tb in the summer season, and t is the time step (1 h). AWD is calculated with a threshold of 18℃ which is lower than minimum summer comfort temperature limit. Thus, AWD18℃ is higher than zero for every climate scenario in which the outdoor air temperature is higher than 18 ℃ for at least 1 h.

Results and discussion
Thermal resilience of the two lecture rooms was assessed for 3 scenarios-(a) long-term climate change (3 TMYs) (b) extreme events (3 severe HWs) and (c) extreme events accompanied by power outages. Existing thermal resilience indicators discussed in section 3.3. were evaluated with FTL and ATL.

Unmet degree hour assessment
4.1.1. Assessment during historical and future TMY period Figure. 6 illustrates the unmet degree.hours for lecture rooms E120 and E220 for 3 TMY periods to assess longterm thermal resilience to overheating. The severity of outdoor warmness is increased by 17% during long-term future(2100s) compared to historical (2010s) TMY period due to higher annual outdoor temperatures. Whereas, the severity of outdoor warmness is slightly reduced (2.5%) during the future mid-term (2050s)TMY compared to the historical (2010s) TMY due to the lower annual average outdoor temperatures in this period.
In both the lecture rooms, without IEC, NNV and shading, percentage of occupied hours above 26℃ (FTL) is higher than 5% acceptable limit. For E120 only in 2100s scenario, the percentage of occupied hours above ATL thresholds are higher than 5% acceptable limit. Whereas for E220, both in 2010s and 2100s scenario, the 5% acceptable limit is violated.
In the lecture rooms equipped with IEC, NNV and solar shading, unmet . ℎ ( with FTL)decreases by 34% during 2050s and increases by 721% during 2100s when compared to 2010s scenario. For both E120 and E220, equipped with IEC, NNV and solar shading, percentage of occupied hours (using FTL) are within 5% acceptable limit. Whereas in 2100s scenario, the 5% acceptable limit is violated. In all 3 TMY period, percentage of occupied hours above ATL are within 3% good limit. Therefore, the selection of the reference thermal comfort model used to estimate overheating is also of paramount importance for the assessment of buildings sensitivity to climate change.

Assessment during historical and future HW period
During the 3 severe HW scenarios (Fig.7), the increase in unmet degree hours is proportional to the severity and the duration of the heatwaves. With and without IEC, NNV and solar shading in both lecture rooms , percentage of occupied hours (using FTL) are > 5% acceptable limit. Without IEC, NNV and solar shading, there is an increase in unmet . ℎ (using FTL) by 812% , 2165% and 72% during severe HWs of 2010s, 2050s and 2100s compared to their corresponding TMY periods. In E120, without IEC, NNV and solar shading there is an increase of 227%, 119% and 195% in unmet . ℎ during HWs compared to the lecture rooms with IEC, NNV and solar shading implemented.
Similarly in E220, without IEC, NNV and solar shading, there is 99%, 78% and 91% increase in unmet . ℎ during HWs compared to the lecture rooms with IEC, NNV and solar shading implemented. During HWs E120 with ATL, percentage of occupied hours are within 5% limit. However in E220, during long-term 2100s severe HW, percentage of occupied hours with ATL is above 5% limit. In both the lecture rooms during the HW accompanied by power outage scenario, 100% of occupied hours are above thresholds (FTL and ATL). Additionally, in case of HW accompanied by power outages, E120 has an average of 18% higher unmet . ℎ than E220. These results demonstrate that for severe short term events like HWs accompanied by power outages, lecture room E220 with medium thermal mass has higher thermal resilience to overheating as they can flush out the heat faster. Whereas, E120 with heavy thermal mass and without passive strategy like NNV, retains more heat and has lower thermal resilience. However in longterm assessment of thermal resilience, lecture room E120 with heavier thermal mass coupled with passive strategies like NNV, has better thermal resilience as it takes more time to absorb heat from outdoor and can flush out heat with passive strategy like ventilative cooling. Similar results were demonstrated in [14].

IOD, AWD and αIOD assessment
Indoor overheating degree will increase as ambient warmness degree increases. In TMY period, IOD for all 3 period is < 1℃. With FTL IOD for both the lecture rooms with and without cooling strategies is less than 1℃. Flores-Larsen et al. [34] extended the study of Hamdy et al. [33] to classify the relationship between heat events and their corresponding IOD. The conclusion from this study proposed thresholds for classifying heat events regarding their impact on the indoor environment: moderate (IOD ≤0.5 •C), strong (0.5 •C < IOD <2.0 •C), and extreme impact (IOD ≥2.0 •C). In TMY scenarios, with FTL, without the cooling strategies in long-term (2100s), IOD exceeds the moderate range to the strong. With cooling strategies, IOD is within moderate range in all 3 TMY scenarios. With and without the IEC, NNV and solar shading, during HW and HW+PO period, IOD >2•C. This is due to the high indoor temperatures during HWs. It can be noted that during the HW of 2050s, IOD is slightly lower than that of 2010s due to variation of HW duration. There is significant increase between 2050s and 2100s (>2•C for HW period and >5•C for HW +PO period. Fig. 9. IOD assessment during 3 TMY period, 3 severe HW period and 3 HW+PO period The severity of outdoor warmness during the 3 TMY and 3 severe HWs is quantified using the Ambient warmness degree. The climate scenarios, namely 2010s, 2050s and 2100s scenario (Fig.10) are represented by their Ambient warmness degrees of 3.51, 3.42 and 4.11, respectively. However, the AWD during the 3 severe HWs is represented by 7.73 (2010s), 9.66 (2050s) and 7.26 (2100s) respectively. The outdoor warmness during the HWs is influenced by the average temperature during the HW ( Table 2).

Fig. 10. AWD assessment for 3 TMYs and HWs scenario
The sensitivity of the building to climate change was assessed using the Overheating escalation factor (αIOD), which quantifies the increase in the Indoor overheating degree (IOD) corresponding to an increase in the Ambient warming degree. Without IEC, NNV and solar shading, αIOD ranges between 0.06 and 0.18 whereas with IEC, NNV and solar shading, αIOD ranges between 0.01 and 0.04. This indicates that two lecture rooms can suppress long-term overheating risk in the future TMY scenarios (αIOD <1). Fig. 11. αIOD assessment with and without IEC, NNV and solar shading for 3TMY scenarios As illustrated in Figure.12, there is significant increase in αIOD when the same building is subjected to thermal shock like HWs and HWs accompanied with power outage. As seen in Figure. 11, with IEC, NNV and solar shading, in respective TMY scenarios with sae duration as HWs, αIOD ranges between 0.1 and 0.13. However, during 3 HWs, max αIOD is 0.41. Thus with cooling, the building is able to supress the overheating risk in HW and in corresponding TMY period. However, even with cooling during a severe shock like heatwave with power outage (cooling strategies shut off during power outage), αIOD range increases between 1.12 to 2.91. This demonstrates that if a power outage accompanies a HW, the building which is able to supress overheating risk during the same HWs, will be subjected to overheating without implementation of cooling strategies. Without cooling strategy, the building is able to supress the overheating risk only in TMY period. During the 2010s HW period, αIOD =1.02. The building is unable to supress the overheating risk. During HW of 2050s (αIOD =0.80) and 2100s (αIOD =0.91), is 97% and 130% respectively higher compared to αIOD with cooling strategy. Without cooling, during a heatwave accompanied by power outage, αIOD is higher than 1 (range between 1.38 and 3.07). During HW accompanied by power outage the difference between with and without cooling is negligible as during power outage the lecture rooms remain without any cooling or shading strategies. Thus without cooling strategies, the building risks severe overheating during HWs accompanied by power outage.

Discussions
The impact of climate change on the thermal resilience to overheating in an educational nZEB in moderate climate is investigated in the current study. Thermal resilience to overheating, is evaluated with existing indicators (unmet degree hours, IOD, AWD, αIOD ) for three TMY climate scenarios and 3 severe HW data based on historical and future projections datasets. An extreme case of severe HWs with accompanied by power outage has also been evaluated. The impact of IEC, NNV and solar shading on the thermal resilience to overheating was assessed. (a) Performance of IEC, NNV and shading: Even though IEC and NNV perform satisfactorily in TMY scenarios, short term thermal shock events like HWs impact the thermal resilience to overheating in buildings. Even with IEC, NNV and solar shading, during HWs there is a significant increase unmet . ℎ in both lecture rooms during HWs compared to period. During HWs and power outages, without the cooling strategies, with both FTL and ATL, the percentage of occupied hours are outside the 5% acceptable limit. However, with implementation of IEC, NNV and solar shading even though the peak indoor temperature improves, yet, with FTL, percentage of occupied hours are higher than 5% acceptable limit. This shows even with passive strategies, there is an overheating risk in existing energy efficient buildings during short-term extreme heat events. This situation is further exacerbates if a HW is accompanied by a power outage. Indoor operative temperatures during day -time remains > 40℃ even 5 days after HWs + power outages are over. This is due to the air-tightness and high insulation in the lecture rooms which retains the heat. (b) Impact of thermal mass: Lecture room with heavier thermal mass demonstrated higher thermal resilience to overheating. However lecture room with heavy thermal mass gets severely impacted if there is an event like HW accompanied by power outage. During short term event like HW heavy thermal mass has higher resistivity to heat gains from outdoors compared to medium thermal mass buildings. On the contrary heavier thermal mass retains indoor heat for longer duration. During power outage as the ventilation strategy is not active, E120 has higher peak temperatures and 18% higher unmet degree hours than E220. (c) Thermal resilience to overheating of the two lecture rooms were studied using three other existing indicators: IOD, AWD and αIOD. The Indoor overheating degree (IOD) will increase as the Ambient warmness degree (AWD) increases. However, the results demonstrate that in TMY period the building is able to supress overheating risk (αIOD<1). In HW period also, thermal resilience to overheating can be improved with the implementation of cooling strategies. This is a similar finding with the unmet degree hour assessment method. During HW period, with cooling and shading strategies the building has good thermal resilience to overheating (αIOD<1). However if a HW is accompanied by power outage, (αIOD>1) and the building do not have good thermal resilience to overheating. However, it is important to point out a few drawbacks of these existing indicators-(1) an indicator like IOD uses the total occupied period instead of the occupied period which are above a certain threshold. This leaves a gap in accessing how many occupied hours in a given time period violate the comfort limit. Although a scale of moderate, severe and extreme impact for IOD has been developed for HW periods, it needs to be further tested on other building types and climate typologies where HW definitions are different (both in terms of severity and duration). (2) αIOD gives an indication of the overheating risk in relation to outdoor warmness. IOD and αIOD indicator still lacks to capture holistic aspects of thermal resilience to overheating assessment. They do not indicate how building or system design parameters when altered will impact the thermal resilience to overheating.

Conclusions and future work
The results concluded that existing energy efficient building with good thermal comfort as seen in the base case scenario have low thermal resilience to overheating during shocks like heatwaves and heatwaves accompanied by power outages. It can also be concluded that thermal mass has significant impact in the thermal resilience to overheating. Medium thermal mass is able to store less heat and thus has lower risk of overheating in long term events whereas heavy thermal mass takes more time to absorb the heat but once stored, it retains the heat and the recovery capacity of the building is low. The study also proved the need for implementation of cooling strategies to improve thermal resilience in future heatwave scenarios. The resilience of the two lecture rooms were assessed with and without IEC, NNV and solar shading. A parametric study to assess the impact of these strategies individually still needs to be done. Furthermore, other passive cooling as well as active cooling strategies needs to be studied.

Acknowledgement
This study is performed under the framework of International Energy Agency's Energy in Buildings and Communities (IEA EBC) Annex 80-Resilient Cooling of Buildings. This work has been supported by the Flanders Innovation and Entrepreneurship in the Flux 50 project 'ReCOver++: Improving resilience of buildings to overheating.'