Evaluation of EN15193-1 on energy requirements for artificial lighting against Radiance-based DAYSIM

This study evaluates the calculation approach of the energy requirements for artificial lighting inside buildings of different use according to EN15193-1:2017, defining the main scope of the standard, highlighting its limitations, and proposing improvements. The evaluation was carried out through a parametric analysis to determine the influence of window-to-wall ratio, distribution of windows, presence of side opening, glazing visible transmittance, and overhang length on the calculation of the Lighting Energy Numeric Indicator (LENI) for a living room and an office at four representative locations (Bratislava, Stockholm, London, Athens). The standard was tested against DAYSIM, a Radiance-based simulation tool for calculating daylight availability, whose results were post-processed to obtain the energy requirements for artificial lighting. For many windows close to each other, the standard’s approach to superimpose the daylight factors (DF) for overlapping daylit areas led to an overestimated total DF and therefore an underestimated LENI. For rooms with low window-to-facade and window-to-wall ratio, the standard’s calculation was inaccurate. The daylight supply factor tabulated in the standard was too low for latitudes below 45°, leading to an overestimation of the LENI. For latitudes above 60°, the opposite effect was observed. Summarising, the standard underestimated the LENI by about 10% on average.


Introduction
Lighting in commercial buildings accounts for up to 45% of overall electricity demand, with significant variation from one building to another [1]. In modern office buildings, electric lighting can provide substantial energy savings with the introduction of reasonable investments [2]. For Northern European countries, it has been demonstrated that the transition to energy efficient lighting systems is one of the most effective and economical methods of reducing CO2 emissions both for new and retrofitted buildings [3].
In recent years, tools for lighting simulation in buildings have become a promising and widely-used method by designers for lighting energy analysis in order to identify the most suitable energy saving options [4]. However, their use is challenging because they require a detailed representation of the real environment. This leads to time-consuming model design and long computation time in case of complex geometries [4]. Moreover, setting up simulations requires very specific knowledge, and the user interface may not be user-friendly [4].

Journal of Building Engineering
The standard EN15193-1:2017 [5] (henceforth referred to as "standard") establishes a calculation method for determining the energy requirements for artificial lighting in buildings. The evaluation is done without creating a comprehensive 3D model of the building, thus permitting fast evaluation during preliminary design. This preliminary information can then be used to inform the actual, detailed design.
The main numerical result from the standard is the LENI (Lighting Energy Numerical Indicator), which quantifies the annual energy consumption for lighting per square meter of treated floor area and is typically expressed in kWh/m 2 ·yr. The detailed calculation method proposed by the standard can be applied for energy certification related to lighting energy consumption of buildings.
The LENI calculation procedure considers, at different levels of detail, the following factors affecting the building's energy consumption for electric lighting: 1) lighting system power, including parasitic power of control systems and power for recharging the emergency lamps; 2) control system type (manual or automatic according to daylight availability, occupancy or both); 3) daylight penetration into the indoor spaces through both vertical glazing and roof lighting systems, which is a function of window-to-wall ratio (WWR), facade window distribution, and glazing visible transmittance; 4) building usage and corresponding lighting requirements, including occupancy time and probability [5].
There are several studies based on the standard's methodology of estimating the annual energy consumption for lighting. For example, in [6] the authors critically discuss the procedure prescribed by the Italian Technical Standards to account for the internal gains in the calculation of the energy performance indices for a building. The paper proposes a new procedure, which relies on the lighting energy numerical indicator (LENI) according to the European Standard EN 15193:2007. The papers [7][8][9][10][11][12][13][14][15][16][17][18][19][20][21] use the approach incorporated in earlier versions of the standardeither EN15193-1:2007 or prEN 15193-1:2015in the conducted tests evaluating daylight availability, energy efficiency and economic benefits from saving potential associated with artificial light energy use. Lo Verso et al. [22,23] present the results of a study quantifying, concerning a manual on/off switch, the energy savings due to the four typologies of daylight-linked controls included in the latest revision of the standard as well as their combination with an occupancy automatic off control. The results show for what combinations of Journal of Building Engineering variables two target savings of 20% and 30% can be reached using the photo dimming and occupancy controls contained in the standard.
From the discussed literature, it can be concluded that the standard's calculation approach is highly recognized and widely used by designers and researchers as the ground truth in the annual energy consumption associated with artificial lighting estimation. For instance, the standard's LENI calculation is included in the software Dialux used for professional light design by designers and manufacturers [24]. The new software LENICALC, developed by ENEA and available since February 2020 , calculates the LENI indicator trying to follow the standard as strictly as possible while guiding the user in setting each required parameter [25].
Nevertheless, few studies exist that compare the standard to more accurate methods [26][27][28][29][30][31][32][33], and most works are related to the previous version of the standard. A major change to the standard's calculation method has been the process of daylight availability estimation [34]. The new version of the standard has been evaluated in [35] in which, based on 108 cases, the authors compared the calculation of the daylight factor (DF) according to the current version of the standard against DAYSIM. The results of the study showed a very good correlation (R2 = 0.99). However, only a single window was included in the model and the case of multiple openings providing daylight to space was not considered.
Additionally, the LENI was evaluated for few of the considered cases, and mainly for comparing different lighting control. Therefore, to better understand the limitations and scope of applicability of the standard, it is necessary to extend the comparisons done in previous studies by considering cases that include important, unevaluated factors, such as a broader range of latitudes, different window-to-floor and window-to-wall ratios (WFR and WWR, respectively), and multiple windows.
This paper provides an assessment of the standard's energy requirements calculation in terms of LENI and DF respectively against DAYSIM [36] and Radiance [37], by performing a parametric analysis that considers, in addition to the evaluation in [35], the impact of the following crucial factors in the calculation of the LENI: window-to-wall ratio, distribution of windows, presence of side opening, Journal of Building Engineering glazing visible transmittance, overhang length, space geometry, and room location. Such an evaluation is important because the new version substantially revised the daylight availability calculation, which has a strong impact on the LENI and therefore on lighting design in buildings. The influence of the above-mentioned factors is in-deep investigated and weaknesses and limitations of the current version of the standard are identified and demonstrated. Finally, suggestions for possible improvements of the standard are given.  Accordingly, the standard's calculation procedure can be divided into the following steps.

Overview of the standard [5]
• Zone under evaluation: definition of its geometry and use; • Estimation of installed electric power from the luminaire types installed in the zone; • Definition of the luminaire control system; • Daylight availability estimation from location (latitude) façade's orientation, weather data, openings geometry, glazing properties, and shading systems. Daylight availability is Journal of Building Engineering a function of daylight factor (DF) calculated by the approach within the areas exposed to daylight; • Result: total energy use and the LENI are calculated. They consider daylight availability and are defined as minimum required electric energy for artificial lighting to meet adequate internal illuminance levels.
The main parameter constituting the LENI is the estimated lighting energy WL,t required to provide a zone of the building with adequate illumination. It is defined by Eq. A2 reported in the Appendix.
WL,t, among other factors, is a function of the daylight dependency factor defined by Eq. A.6.
depends on building location and geometry along with openings geometry, obstructions presence, shading device presence/absence and glazing properties.
One of our goals was to compare daylight availability calculated according to the standard with a calculation in DAYSIM. We did not evaluate the need for solar shading activation due to glare.
Therefore, we neglected the factor called "solar/glare protection activation" [5], accounting for glare protection (see Eq. A.9), i.e., we assumed no internal shading. Thus, Eq. A.9 becomes: , , = , , , , denotes the daylight supply factor of surface j evaluated whenever the solar protection system is inactive. It is a function of the site latitude γ, the ratio Hdir/Hglob (so-called luminance exposure) between direct (Hdir) and global (Hglob) illuminances calculated on the horizontal plane, façade orientation, level of maintained illuminance (Em) and DF [5]. The daylight supply factor , , , is a tabular value in the standard.

Evaluation method
The evaluation is based on the parameters that constitute the calculation steps defined by the standard.

Parametric analysis
The main changes made in the standard's recent revision [32] concern the process of daylight availability estimation within the energy assessment. Therefore, we focused ourselves on parameters Journal of Building Engineering regarding the standard's calculation of daylight availability, which are essential for calculating the LENI.
In order to estimate the energy use for artificial lighting, the standard derives a quantity termed "daylight factor" (DF) that accounts only for areas exposed to daylight. However, the commonly accepted definition of the DF (see, e.g., [38]) is different, because it is evaluated on a different reference area, as discussed in Section 3.4. Therefore, we also investigated the appropriateness of the DF calculation as per standard with the use of Radiance because it is essential for understanding the reasons for differences in the LENI results compared with DAYSIM.
Accordingly, the factors chosen for the parametrical analysis are: • because the standard provides tables for these locations and to assess whether the accuracy of the results depends on the climatic zone. Table A.1 shows the geographical information and luminous exposure Hdir/Hglob of the locations. Building location influences the LENI, but not the DF calculation.

• Room dimensions
Two different room dimensions were chosen based on the sample geometries from [39,40]. The dimensions are listed in Table 3.1. The dimensions in Table 3.1 refer to the centerline of the walls. Since within the standard approach the thickness of elements that constitutes the zone is neglected, in our simulations, the walls, floor and ceiling were modeled with zero thickness.
Both the living room and the office are south-oriented, meaning that the window openings are Journal of Building Engineering located on the south façade for both building uses.
Zone geometry influences the daylit area and therefore affects both the DF and the LENI calculation.

• South facade window-to-wall ratio
The window-to-wall ratio (WWR) is parameterized by window height and sill level above the floor as shown in Table 3.2: The current version of the standard does not explicitly consider the window frame. In the standard, WWR is understood as the transparent (glazing) over the opaque (including the window frame) part of the façade. Therefore, we modeled all windows in DAYSIM and Radiance without a frame to be compliant with standard's WWR definition in our analysis.
WWR is in direct relation with WFR (window-to-floor ratio), which is the ratio between the transparent window area (glazing) over the analyzed space floor area. The dependence between WWR and WFR for the geometries that we used for our analysis is shown in Table 3.3:

Glazing visible transmittance
Double and triple glazings were simulated with a visible transmittance of 0.73 and 0.63, respectively [41]. •

Horizontal overhang presence
The presence of an overhang in the calculation of the standard influences light penetration and affects energy consumption.

Journal of Building Engineering
Three different depths for the horizontal overhang were tested: 0.2, 0.4 and 0.6 m. The overhang was placed at the upper edge of the window glazing. The case without overhang has also been considered.

Design of experiment
We performed two full factorial analyses, one for the DF and one for the LENI, using the factors listed in Section 3.1.
As already mentioned, the DF estimation as per standard was compared with Radiance results because the LENI calculation in the standard depends on the DF.
Since the DF is independent of building location and window orientation, the full factorial design is given by (levels in parentheses): building geometry (2)

Parameterization tree for assessing DF estimation
The LENI parameterization includes additional factors such as location (4) Table 3.4 shows two renders as an example of modeled geometries:

Task plane, luminaries and control system
The height of the task plane is fixed at 0.8 m above the floor level, which is assumed to be the height of the working desk in [5].
The luminaire chosen for all calculations is an LED lamp with a constant illuminance dependency factor Fc = 0.85, a maintenance factor MF = 0.7, and a light source efficiency factor FL = 0.86 [5]. The luminaires were set at the ceiling level for both geometries, which is 2.5 m above the task plane for the office and 1.9 mfor the living room. Upward flux fraction was set to 30%. Following the Eq.A.3 the lighting power densities were calculated: 14.09 W/m 2 for office and 15.59 W/m 2 for the living room.
Daylight responsive control as modeled in the standard considers imperfections of control systems, such as time lags for the activation and deactivation of lighting, inaccuracies in measuring the illuminance level, and manual lighting control by the occupants. However, the standard is not explicit enough about these imperfections to allow for implementation in simulation software. Therefore, we decided to simulate only a simple control logic that assumes that lighting is on if the daylight Journal of Building Engineering illuminance on the working plane is below 500 lux, and off otherwise. The 500 lux threshold value was set for both the living room and office geometries.

Daylight factor assessment
The DF is normally calculated considering the whole area of the room under evaluation [38]. In contrast, the standard proposes a simplified calculation of a quantity termed "daylight factor" that only considers areas exposed to daylight. It is then used for the energy requirements calculation within the  2+4=6 [41] The overall DF (as per standard) for the total daylit area of the zone is found as a sum of weighted averages of the DFs for each daylit floor patch. The weight is given as a daylit floor patch area divided by the total daylit area. Eq. 3.1 demonstrates the calculation of the final daylight factor of the zone as per standard.

Journal of Building Engineering
Here, DFi is the daylight factor calculated on the single daylit area Ai.
Each single daylit area is estimated as: Where is the width of the window [m] and ad is the depth of the daylit area [m] (see Fig. 3.5).

Fig. 3.5 Estimation of single daylit area with maximum depth admax, window lintel height hLi [m]
and task area height hTa [m] [41] We used a sweep line algorithm to calculate the total daylit area [42].

Software used to evaluate the standard
As a reference for the standard evaluation, the Radiance-based software DAYSIM v4 [6] was used.
DAYSIM is a validated [36,43] and highly recognized by professionals software in the field of daylighting design and verification [4].
DAYSIM calculation is based on the daylighting coefficient method [44], which allows to carry out a fast annual daylighting simulation. DAYSIM uses the records of direct normal and diffused horizontal irradiances and feeds them into the Perez All-Weather model [45], which is composed of a model that derives illuminance values from irradiances and models that recreate a luminance distribution on the sky vault from illuminance values [46]. The "Interpolated method" of DAYSIM was used in the The annual energy use for artificial lighting was derived from illuminances using a Python v3.7.1 [53] script for post-processing. The script is described in Section 3.6.

DAYSIM model definition
Test points for illuminance estimation were generated on the workplane situated at 0.8 m height [5]. The DAYSIM ambient parameters set for the case studies are listed in Table 3.6. parameters: ab (ambient bounces) and lw (limit weight), while aa (ambient accuracy), ad (ambient divisions) and ar (ambient resolution) were set to be accurate [54]. A detailed explanation of the effect of the parameters is reported in [55].

Journal of Building Engineering
The convergence test was performed considering the following indicators: the LENI and average Daylight Autonomy with a threshold of 500 lux (DA_500) [56]. We considered an example case with the following parameters: location Bratislava, office (corresponding schedule described below), WWR 0.5 on south-facing façade with 1 window, side opening on the west facing façade, triple glazing, no overhang.  Journal of Building Engineering

Fig. 3.7 Dependency of the LENI and DA_500 on -ab in our example case, -lw is set to 0.01
Results show that from ab from 5 to 8 the LENI and DA_500 remain stable. Comparing the results for the two -lw settings, it can be observed that the LENI and DA_500 output are similar. Therefore, ab was set to 5 and -lw to 0.01.
We assumed that the luminaires were fully switched on (no dimming) if the average daylight illuminance inside the test room was below the established 500 lx.
The use schedules were set as follows.
Our Python script extracts illuminances within these schedules and applies the following condition: if the mean daylight illuminance calculated on the task area (0.8 m above the floor level) is lower than 500 lx, artificial light is turned on. By doing this, we obtain the annual number of hours when the artificial light is on.
According to Eq. A.5, the LENI is affected by daylight time when lighting could be used and daylight absence time . These parameters depend on the latitude, and their calculation procedure is Journal of Building Engineering provided by the standard. Knowing the schedule, tD and tN were calculated from weather data [51] using a custom-made Python script. Any hour with an illuminance greater than zero was considered daylight time, otherwise daylight absence time. The data is listed in Table 3.7 and was used in both the DAYSIM and EN15193-1:2017 models.

Results and discussion
Results are subdivided according to the parametrization trees for DF and the LENI estimation (see Section 3.2).

DF evaluation
As discussed in Section 3.4, in the approach for the LENI calculation adopted by the standard, the DF is estimated only on areas exposed to daylight. This is different from the conventional way of   The largest differences between the standard and Radiance results are found for the south WWR 0.7.

Journal of Building Engineering
From south WWR 0.5 to 0.7, Radiance results change slightly whereas there is a significant increase in DF as per the calculation of the standard.
This divergence can be explained by the fact that in both geometries (office and living room), changing the south WWR from 0.5 to 0.7 and keeping the same number of windows, the height of the windows remains the same. The only parameter that changes is the windowsill level, as illustrated in  If the windowsill is below the task area, only the reflected light by the internal surfaces of the space can additionally increase the illuminance level on the task area. The Radiance results, therefore, make more sense in this case, keeping almost the same illuminance level for both WWR 0.5 and 0.7. For WWR 0.7, the standard overestimates the DF by adding the direct light from the part of the window below the task plane to the illuminance on the task plane. It is done due to the mathematical representation of the standard, since the procedure (Eqs. A.6 and A.7) considers the window area. Splitting window 1* into two windows of 0.35 m width each with zero distance between them, thus in the same position and of the same total size as window 1*, the configuration on the right of Fig. 4.5 is obtained.

Fig. 4.5 Discrepancy in DF results as per standard because of daylit area superposition
The DF for the case on the right of Fig. 4.5 is calculated as per standard as follows: • Total daylit area (the same as daylit area 1*) is 6.07 m 2 (calculated with a sweep-line algorithm [42]); • Daylit area from opening 1 is 5.136 m 2 (the same as in simulation no. 5), which results in a weighting coefficient of 5.136 / 6.07 = 0.846; • Daylit area and weighting coefficient from opening 2 are the same as for opening 1; • Overall DF is 1.823 × 0.846 + 1.823 × 0.846 = 3.08 (35% overestimation compared with the case on the left of Fig. 4.5).
This example demonstrates that the superposition principle proposed by the standard for the evaluation of the overall DF leads to an error in some cases. Possible refinements of this approach should be investigated.
This overestimation of the overall DF leads to an underestimation of the LENI for a high number of windows on the south façade, as shown in Section 4.2.
Journal of Building Engineering

LENI evaluation
For easier readability and comparison, the simulations were subdivided into 16 groups according to the parameterization in Fig. 3.3, i.e. four locations, two room types and the presence or absence of a window on the west facade. In the following sections, each location is analyzed in a separate subsection.  The calculation as per standard overestimates the LENI significantly in most cases. This is mainly due to the calculation of the daylight supply factor , , and daylight dependency factor , see Eqs.

Athens
2.1 and A.8, as explained in the following.
According to Eq. 2.1, the daylight supply factor is equal to , , , . This value is tabulated in the standard.
To explain the observed differences, we provide a calculation example for the office geometry with a west facade WWR of 0.    [58], hence it is expected to benefit from more daylight than the other locations. This is confirmed by the luminous exposure, which is the highest for Athens, see Table   A.1.
The daylight dependency factor FD for each location is reported in  This observation along with the LENI evaluation in Fig. 4.6 raises the question of whether the standard's tables for the evaluation of the daylight supply factor are reliable for southern latitudes.   Table A.1).This is why the daylight dependency factor FD is almost the same for both locations (see Table 4.1) in the example calculation in Section 4.2.1. We conclude that Journal of Building Engineering the tabulated values for , must be less appropriate for Stockholm than for Bratislava and lead to an overestimated daylight dependency factor and therefore an underestimated LENI.

Effect of the overhang and overall LENI evaluation
As reported in the previous sections, discrepancies were found to be larger in the presence of an overhang. Fig. 4

Conclusions
This study presented an evaluation of the LENI and DF calculation according to standard EN15193- In the following, we summarize the issues found in the standard's calculations and provide suggestions for improvements. This assessment then allows us to report the range of applicability of the standard.

Summary of validation investigation, suggestions for improvement
1. For rooms with low WFR and WWR the standard's calculation is inaccurate.
2. In the DF calculation, the standard does not consider the windowsill level but only the window area. Thus, the standard cannot discern whether the task area is situated above or below the windowsill. However, this is crucial knowledge to adequately determine the illuminance over the task plane. By directly adding the contribution of direct light from the part of the window below the task plane to the illuminance on the task plane, the standard overestimates the DF and therefore underestimates the LENI.
3. For a high number of windows close to each other, the standard's approach to superpose the DFs for overlapping daylit areas leads to an overestimated DF and therefore an underestimated LENI. 4. The values for the daylight supply factor tabulated in the standard appear to be too low for latitudes below 45°. This leads to an overestimation of the LENI.
5. For latitudes above 60°, the daylight supply factor reported in the standard appears to be Journal of Building Engineering overestimated, which leads to an underestimation of the LENI.
To address the first two issues, our suggestion is to use Radiance or other validated tools instead of the standard for the calculation of the DF whenever several windows extend below the task plane or are close to each other. Further research is needed to identify suitable quantitative criteria. To address the third issue, a lower limit on WFR should be introduced into the next version of the standard. Possible thresholds are found in design codes. The fourth and fifth issues could be fixed by revising the respective tables in the standard.
Results for Bratislava and London demonstrate that the standard's calculation is adequate for locations in central Europe (i.e., latitudes between 45° and 60°). In this range, the standard's calculation is more consistent with DAYSIM results for cases with higher DF and therefore lower LENI. The LENI is also more consistent in cases with a side opening (which leads to a higher DF).
As a general conclusion, the standard tends to underestimate the LENI except for the case of Athens with the lowest LENI values. The underestimation is more pronounced for higher values of the LENI.
Practitioners should keep this in mind when applying the standard, especially because they might improperly assume that a simplified procedure such as the one proposed by the standard would overestimate the LENI to provide a conservative result.

Scope of applicability
It can be concluded: to obtain results of better reliability, it is suggested to apply the EN15193-1:2017 approach to spaces designed in the central European locations (latitudes within the range of 45°-60° with higher values of luminous exposure -around 0.5), with WWR's modeled in the range 0.3 -0.5, keeping in the account that the sill level of openings is above the task area. The better results are achieved in cases where more than one façade has openings on itdue to daylight supply factor estimation tables.
In this sense, the addition of shading device of various size and form along with linear obstructions, as well as the reduction of visible transmittance values due to the use of different glazing typologies, and all the other factors that may influence the daylight penetration within the building, can contribute Journal of Building Engineering to the underestimation of LENI results with a different sensibility. Generally, the bigger the parameter change affects the DF results, the lower is the reliability of LENI results.
In the authors' opinion, it is important to keep in mind that generally, EN15193-1:2017 tends to underestimate LENI results.