Dynamic Daylight Metrics for Electricity Savings in Ofﬁces: Window Size and Climate Smart Lighting Management

: Daylight performance metrics provide a promising approach for the design and optimization of lighting strategies in buildings and their management. Smart controls for electric lighting can reduce power consumption and promote visual comfort using different control strategies, based on affordable technologies and low building impact. The aim of this research is to assess the energy efﬁciency of these smart controls by means of dynamic daylight performance metrics, to determine suitable solutions based on the geometry of the architecture and the weather conditions. The analysis considers different room dimensions, with variable window size and two mean surface reﬂectance values. DaySim 3.1 lighting software provides the simulations for the study, determining the necessary quantiﬁcation of dynamic metrics to evaluate the usefulness of the proposed smart controls and their impact on energy efﬁciency. The validation of dynamic metrics is carried out by monitoring a mesh of illuminance-meters in test cells throughout one year. The results showed that, for most rooms more than 3.00 m deep, smart controls achieve worthwhile energy savings and a low payback period, regardless of weather conditions and for worst-case situations. It is also concluded that dimming systems provide a higher net present value and allow the use of smaller window size than other control solutions.


State of the Art
Energy saving is one of the key variables in present-day building construction and civil engineering. In fact, lighting represents between 15% and 30% of power consumption in buildings [1][2][3]. Accordingly, suitable use of daylight is essential in reducing the power consumption of electric lighting [4], while the development of new technologies, such as the improvement of LED lamps or lighting smart controls, can help promote a lower impact on the environment [5].
Lighting smart controls were introduced in the early 2000s to promote energy saving in buildings. One of the first outstanding examples of their use in buildings is the New York Times Headquarters, where a lighting dimming system based on occupancy and daylight availability achieved an energy saving of close to 40% in the floor perimeters [6]. Subsequently, lighting smart controls were used in other buildings of note [7,8] with a noticeable energy saving in lighting.
considering a variable reflectance of the inner surfaces as well as a diffuse reflection. Thus, the light reflected is directly proportional to the cosine of the angle between the observer's line of sight and the surface normal. The different sizes of openings in the façade are defined as surface ratios. The window opening is double glazed with a visible light transmittance of 0.70 and 0.05 m thick joinery. The virtual room and calculation variables are shown in Figure 1 and Table 1.
The study points at which dynamic daylight performance metrics were analyzed were positioned on equidistant axes at a height of 0.70 m with a spacing 0.75 m wide and 0.25 m deep, as shown in Figure 1.  Figure 1 and Table 1.
The study points at which dynamic daylight performance metrics were analyzed were positioned on equidistant axes at a height of 0.70 m with a spacing 0.75 m wide and 0.25 m deep, as shown in Figure 1.  Table 1 shows 18 room models established according to a variable depth (values of 3, 6 and 9 m), window size (window-to-façade ratio from 30 to 90%) and the reflectance of the inner surfaces.   Table 1 shows 18 room models established according to a variable depth (values of 3, 6 and 9 m), window size (window-to-façade ratio from 30 to 90%) and the reflectance of the inner surfaces.  A variable width (values of 4, 8 and 12 m) was considered to assess the energy savings according to the room surface, producing a total of 54 room models. This variable barely affects the calculation results of the dynamic metrics and is therefore only used to determine the cost-effectiveness metrics used to validate the lighting smart control from an economic perspective.

Location of the Room Model
The room models defined were studied for three different locations-Stockholm, London and Madrid-representing a wide range of weather conditions and latitudes from 40 to 60 degrees, thus contributing to the analysis of the impact of latitude and sky luminance. Accordingly, the results obtained for Madrid could be extrapolated to the Mediterranean climate, while the conclusions for London and Stockholm could be assumed for other parts of Northern Europe. The weather data for these three locations were obtained from EnergyPlus Engineering Reference [25], using direct normal and diffuse horizontal irradiances, as well as from the sky model developed by Perez et al. [26] and accepted by CIE [27]. The files selected for Stockholm and London, STOCKHOLM-ARLANDA IWEC and LONDON-GATWICK IWEC (International Weather for Energy Calculations), were created and provided by the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) [28]. The file selected for Madrid, MADRID SWEC (Spanish Weather for Energy Calculations), was created using data from Pérez-Lombard at the Spanish National Institute of Meteorology (AEMET) [29].

Orientation of the Window
All the windows in this study face north, avoiding direct sunlight since this is the worst-case scenario for indoor daylight illuminance values [30]. In office buildings facades facing north do not usually have blinds [31,32], since fundamentally their use can be limited to the control of the glare in the initial and final hours of the day, in many cases outside the working hours. The influence over the control of the annual gain of natural lighting is usually of little relevance [33]. This also allows a better comparison between locations. Despite the fact that the height of the lintel affects daylight penetration, it is not considered for this study, nor are the neighboring solar obstructions. The ground reflectance value used is 0.2, which is the default value recommended by DaySim and provided by the Radiance engine.

Lighting Design of the Room Model
The electric lighting of the virtual room consists of the use of Erco Skim downlight oval LED floodlight luminaires (Figure 2). These 37 W luminaires, with a luminous flux of 3690 lm, are arranged in parallel lines following the depth of the room: a row of luminaires is placed every 3 m deep (one, two or three rows, depending on the depth of each model), beginning at 1.5 m from the window.
This study considers an illuminance threshold of 500 lx, standard value for offices according to EN 12461-1:2012 [34]. Given that variations in the threshold cited would affect the results on the impact of energy efficiency from lighting smart controls, the illuminance value must be chosen carefully, based on use. The spacing between luminaires was calculated using Dialux 4.12, a simulation program validated by previous studies [35] and widely used in electric lighting design. In accordance with the results obtained, the separation between luminaires in each of the rows was optimized to achieve a uniformity value above 0.6 and an average illuminance of 500 lx, obtaining a spacing of 1.25 m, as can be seen in Figure 2.
To obtain a value close to the illuminance threshold, the luminaires are initially dimmed to reach 500 lx on the work plane, thus adjusting electricity consumption. The luminous flux adjustment of the lamps corresponds to 75% in those rooms with high reflectance values and to 90% for dark rooms. Defining the minimum luminous flux for each study case, the minimum energy savings promoted by lighting smart controls can be quantified.
For the sake of comparison and to keep the model as simple as possible, the base case (Case study 0) will be wired to only one command circuit. Although in this type of room it is usual to have at last two (or three) switches, it is not uncommon to find the simple all-on/off situation-mainly in older buildings. Three alternative case studies were developed (Figure 2) based on the typical control approaches for daylight saving [36]: • Case study 1: Manual On/Off lighting control with two separate control rows: circuit 1 is for the near-façade lighting row and a second command control is for the remaining lighting rows. This system is only available for rooms 6 and 9 m deep. • Case study 2: Common Dimming lighting control for all the luminaires of the room (single controller). The dimmer is controlled by a lux-meter which detects daylight illuminance, adjusting the power supply for the lamps. • Case study 3: Two independent dimming-lighting-controls with two separate groups, where one circuit commands the near-façade lighting line and the other the remaining lighting rows. Systems are controlled by lux-meters which detect daylight illuminance, adjusting the power supply for the lamps. This system is only available for rooms 6 and 9 m deep.
It is worth noting that the proposed luminaires only determine the power consumption in electric lighting, following optimal location and photometric distribution. According to the results of this brief study, the electric cost, and in turn the suitability of the proposed smart controls, can be defined by their initial investment costs.
To determine the minimum energy saving of the dimming smart controls, it is assumed that occupants will not delay the activation of the switching controls to achieve the threshold of 500 lx (acting as perfect users [37]). Accordingly, the automatic response of the occupants serves to determine the tightest baseline scenario for quantifying the minimum energy savings achieved by the lighting smart controls. It is foreseeable that the savings will be higher the farther away these users are from the perfect user. It should be noted that this approach is based on daylight availability and on an ideal functioning of the control system, the response of which is actually affected by the calibration setting, the photo-sensors' characteristics and the actual daylight fluctuations, as demonstrated in previous research [38,39]. Therefore, further research is necessary in order to evaluate energy savings in a more realistic operative mode.
Moreover, as the location of the illuminance-meters affects the dimming control, these are located in the central axis of the room, 3 m from the façade (case 3) and at the back of the room near the inner wall to assure the threshold in all circumstances (worst case scenario). The power consumption of other LED lamps would be very similar to those chosen for this study, obtaining a value of energy efficiency in lighting close to 1.5 W/m 2 /100 lx for bright rooms and 1.7 W/m 2 /100 lx for dark rooms. In the case of halogen or fluorescent lamps, the energy efficiency would be noticeably poorer, consuming more energy than the luminaires selected. Therefore, LED lamps represent the most conservative scenario for calculating the suitability of smart controls.
Since this study evaluates the initial investment costs of the proposed smart controls, Table 2 shows the economic costs of case studies 1, 2 and 3, including electric components, dimming control system, wiring and assembly costs. The total costs listed exclude the costs of the components of Case study 0, since these elements are common to all the case studies and can be considered a reference.  The power consumption of other LED lamps would be very similar to those chosen for this study, obtaining a value of energy efficiency in lighting close to 1.5 W/m 2 /100 lx for bright rooms and 1.7 W/m 2 /100 lx for dark rooms. In the case of halogen or fluorescent lamps, the energy efficiency would be noticeably poorer, consuming more energy than the luminaires selected. Therefore, LED lamps represent the most conservative scenario for calculating the suitability of smart controls.
Since this study evaluates the initial investment costs of the proposed smart controls, Table 2 shows the economic costs of case studies 1, 2 and 3, including electric components, dimming control system, wiring and assembly costs. The total costs listed exclude the costs of the components of Case study 0, since these elements are common to all the case studies and can be considered a reference.

Parameters of the Calculation Program
The lighting simulation program used to determine the dynamic daylight metrics and the energy saving in electric lighting is DaySim 3.1. This software is based on the Radiance engine, developed by the Building Technologies Department at the Lawrence Berkeley National Laboratory, and validated by several studies [40,41]. DaySim was designed to achieve a more accurate calculation than the initial form of Radiance, defining the current metrics according to modern sky definitions [26]. Like the Radiance engine, DaySim has been validated by several researchers [42,43] using CIE test cases [44]. Table 3 shows the calculation parameters used by this program in this study. An interval of 5 min is considered for the measuring of the illuminance values during the calculation period. The illuminance requirements and the occupancy hours are described below.

Daylight Metrics and Conditions
Two dynamic metrics were assessed in the room models defined above. The first of these was daylight autonomy (DA), a concept conceived by the Association Suisse des Electriciens [45] and redefined by Reinhart et al. [46]. This metric is defined as the percentage of the year when a minimum illuminance threshold is met by daylight alone so that the higher the daylight autonomy, the lower the power consumption in electric lighting. This metric can be defined as Equation (1): where DA is daylight autonomy, t i is the occupied time in a year, w f i is the weighting factor which depends on the illuminance threshold, E D is the daylight illuminance measured at a given point, and E L is the illuminance threshold. The second dynamic metric is continuous daylight autonomy (DAC) which represents the percentage of the year when a minimum illuminance threshold is met by daylight alone, considering a partial credit linearly to values below the threshold defined [47]. Therefore, this metric can be expressed as Equation (2): where DAC is continuous daylight autonomy, t i is the occupied time in a year, w f i is the weighting factor which depends on the illuminance threshold, E D is the daylight illuminance measured at a given point, and E L is the illuminance threshold. According to the previous formulae, the dynamic metrics are calculated depending on the weather conditions which define daylight illuminance, the illuminance threshold and the occupancy time. The three locations selected for this study represent a wide range of weather conditions and latitudes from 40 to 60 degrees. As with the electric lighting design, the illuminance threshold is 500 lx, a standard value for offices according to EN 12461-1:2012 [34]. Finally, occupancy begins at 8.00 am and finishes at 5.00 pm, following the typical schedule for office rooms.
As can be deduced from the metrics described above, the assessment of daylight autonomy can determine the percentage of use of the lighting system in this time frame and thus, the power consumption in electric lighting using an On/Off control system. Moreover, the analysis of continuous daylight autonomy can ascertain not only the on-time of the electric lighting, but also the amount of light provided when it is turned on and thus, the power consumption in electric lighting when a dimming system is in place. Both metrics therefore are useful in determining the energy efficiency produced by the smart controls proposed in this study.

Cost-Effectiveness Metrics and Conditions
This study uses the net present value (NPV) indicator to evaluate and compare the economic profitability of the proposed lighting control hypotheses. This metric establishes the economic return on the investment for a given number of years [48], as the following expressions (Equations (3)-(5)) show: where: • I 0 is the Initial Investment cost, shown in Table 2  The flow of benefits (FB i ), affected by the national price of electric energy and its fluctuations, is difficult to predict, especially when a period of 10 years is considered (NPV 10 ). This is because the Annual electric energy cost used here was that of 2016 and the Annual Growth Rate of the electric energy cost (AGR E ) was calculated from the annual electricity prices from 2005 to 2016 for each country under study, according to EUROSTAT [51]. Both EC and EC i for each country under study are shown in Table 4: Thus, the economic viability of the investments performed can be evaluated according to NPV value; the higher the NPV, the better the return on the lighting control system. It is also interesting to note the year in which the NPV value changes from negative to positive, as this shows when the investment starts to yield profits. This time indicator is the payback period (PP) which is assessed in this research to determine the suitability of the smart controls proposed.

Validation of the Calculation Program and the Dynamic Metrics
The calculation program and both dynamic metrics were validated, given that a computational simulation is not reliable until it has been compared to a real model. For this purpose, an existing test cell, located in Seville (Spain) was used as a reference [52].

Characteristics of the Test Cell for Validation Process
The room selected for this validation process is one of the test cells of TEP-130 research group [53], located in Seville (Spain) and facing south, in order to optimize rehabilitation solutions on façades and windows in the Mediterranean area.
The real model, which generates the predictive results for DA and DAC, is defined from the test cell characteristics, as a room 2.40 m wide by 3.20 m deep by 2.70 m high. The entire enclosure, including the floor and the roof, is built using high density sandwich panels with a combined thickness of 460 mm, colored in white and screwed to a steel frame structure. The wall facing south has a window 116 cm wide by 108 cm high, with aluminum sliding frame and double glazing (two 4 mm glass and an 8 mm air space) with a solar factor of 0.75. A conservation factor of 0.8 is considered for The illuminance values for DA and DAC indicators were measured during 2017, from 1 January to 31 December (one full year), with eight illuminance meters (range 20-2000 lx, accuracy ±3.0%) placed on the axis of symmetry of the room spaced 0.40 m apart and 0.06 m above ground level, as can be seen in Figure 3a.

Calculation Model for Validation Process
The calculation model has been defined following the geometry and characteristics of the test cell described above, as seen in Figure 3b, considering the same measures and reflectance values for the inner surfaces. As in the case of the test cell, the calculation grid of the virtual model represents the location of the illuminance meters above the floor. The calculation parameters used for this virtual model are described in Table 3.

Calculation Model for Validation Process
The calculation model has been defined following the geometry and characteristics of the test cell described above, as seen in Figure 3b, considering the same measures and reflectance values for the inner surfaces. As in the case of the test cell, the calculation grid of the virtual model represents the location of the illuminance meters above the floor. The calculation parameters used for this virtual model are described in Table 3.

Calculation and Measurement Conditions of Validation Process
The calculation of daylight autonomy (DA) and continuous daylight illuminance (DAC), both for computer simulation and measurements, considered occupancy hours from 8:00 a.m. to 5:00 p.m., with no break for lunch or blind control. Given that DA and DAC depend entirely on indoor illuminance values, the illuminance threshold variable for the calculation has three values-100, 250 and 500 lux-representing the average illuminance range recommended in most common uses of architectural spaces. Table 5 shows the dynamic daylight metrics measured at the study points for the defined illuminance thresholds of 100, 250 and 500 lux, both from annual measurements and dynamic simulations. This table also shows the divergences between measurements and simulation, expressed in percentages.

Calculation and Measurement Conditions of Validation Process
The calculation of daylight autonomy (DA) and continuous daylight illuminance (DAC), both for computer simulation and measurements, considered occupancy hours from 8:00 a.m. to 5:00 p.m., with no break for lunch or blind control. Given that DA and DAC depend entirely on indoor illuminance values, the illuminance threshold variable for the calculation has three values-100, 250 and 500 lux-representing the average illuminance range recommended in most common uses of architectural spaces. Table 5 shows the dynamic daylight metrics measured at the study points for the defined illuminance thresholds of 100, 250 and 500 lux, both from annual measurements and dynamic simulations. This table also shows the divergences between measurements and simulation, expressed in percentages.   As can be deduced from Table 5, daylight autonomy (DA) values are close to those observed in simulations, with a maximum deviation of 8.3% for 250 lux and 8.4% for 500 lux, respectively. These differences show a small and progressive divergence between measurements and simulations in relation to depth, but they can be considered acceptable due to the low values for all the illuminance thresholds.

Analysis of Validation Process Results
In the case of continuous daylight autonomy (DAC) values, divergences are smaller than for DA, with a maximum deviation of 4.2% for 100 lux, but coinciding more at higher illuminance thresholds.
The bias error for both metrics is 1.9% for DA and 1.0% for DAC, with a standard deviation (95% reliability) of 6.8% and 4.9%, respectively. In both cases, these divergences are below 10% and are therefore acceptable.
From the analysis and results obtained, it is concluded that DaySim 3.1 provides an accurate calculation of dynamic daylight metrics.

Quantification of Power Consumption in Stockholm
Following the methodology defined above, Figure 4 shows the sections for all room models, displaying the dynamic daylight metrics and the average power consumption measured at the central axis for each type of control system, based on the results obtained for the Stockholm location. The first column represents rooms 3.00 m deep, the second rooms 6.00 m deep and finally, the third displays rooms 9.00 m deep. Moreover, the first and second rows show the rooms with a window-to-façade ratio of 30%, the third and fourth rows rooms with a window-to-façade ratio of 60%, and the last two rows rooms with a window-to-façade ratio of 90%. Odd rows represent the bright rooms while even rows show the dark rooms. The identifier for each room, defined in Table 1, is on the upper-right side of the section.
As stated earlier, daylight autonomy can determine the average power consumption in electric lighting using an On/Off control system, given that this metric defines the percentage of the year when the threshold of 500 lx is achieved by daylight alone, establishing the turn-on time of the luminaires. Moreover, continuous daylight autonomy determines the average power consumption in electric lighting with dimming controls. Therefore, the average power consumption is shown for each luminaire row, considering On/Off and dimming controls.
Continuing with the results obtained, the average power consumption of the lighting smart controls defined in the methodology is given in Table 6, reflecting electric consumption in W/m 2 according to case studies 0 (On/Off control for all luminaires), 1 (On/Off control with two separate lines), 2 (Dimming control for all luminaires), and 3 (Dimming control with two separate lines).
It is also worth noting that the control system used in Case study 0 (an auto-switching system with one zone) affects the savings of the deepest models. For example, for a 9 m deep space, the savings decrease as the window area increases due to the fact that daylight savings are obtained in the Case study 0, and these are greater for large windows that would permit the entire room to be switched off more often.
As seen in Table 6, deeper rooms require higher power consumption, since daylight cannot reach the back of the room and the dependence on electric lighting is therefore high. This dependence is higher for dark rooms, with low reflectance of the inner surfaces, given that the reflection of daylight could contribute to an increase in illuminance and a reduction in the turn-on time of the luminaires.
Moreover, it is observed that window size has a notable effect on power consumption. Except for deep rooms with low reflectance, the largest windows (window-to-façade ratio of 90%) account for between 35% and 50% less power consumption in lighting than small windows (window-to-façade ratio of 30%). This energy saving is lower for medium windows (window-to-façade ratio of 60%) which consume between 10% and 30% less power than small windows.    Finally, smart controls can reduce power consumption in electric lighting. Dimming controls specifically produce energy savings close to 30% compared to the conventional On/Off controls and the systems that control the luminaires in separate rows save up to 20% of energy. Combining both strategies, the dimming controls of separate rows of luminaires (Case study 3) can save between 35% and 55% compared to the typical On/Off control (Case study 0).
Based on the results in Figure 4 and Table 6, the annual energy saving is summarized in Table 7, depending on the room dimensions, window size, surface reflectances and smart controls proposed. The annual energy saving is obtained by comparing the average power consumption of case studies 1 (On/Off control with two separate rows), 2 (Dimming control for all luminaires), and 3 (Dimming control with two separate rows) to the typical On/Off control, defined as Case study 0. As Table 7 shows, energy saving is higher for cases with small windows which depend more on electric lighting. This rule does not apply to deep rooms, because the conventional On/Off control (Case study 0) is almost always on regardless of window size, and the smart controls save more energy in this case, even more so with large windows.
Moreover, it is worth noting that the use of smaller windows, which may consume more energy in electric lighting, can make up for this weak point using smart controls. For example, a room with a medium window (window-to-façade ratio of 60%) can produce higher energy savings than a room with a larger window if a dimming control with two separate rows (Case study 3) is used. Figure 5 shows the cross sections for all room models for the London location, together with the dynamic daylight metrics and the average power consumption at the central axis. This figure has a similar structure to the previous one, defining the depth of the room in the columns and the window size and reflectance of the surfaces in rows. As in the calculation above, the identifier for each room, defined in Table 1, is on the upper-right side of each section. Figure 5 shows the cross sections for all room models for the London location, together with the dynamic daylight metrics and the average power consumption at the central axis. This figure has a similar structure to the previous one, defining the depth of the room in the columns and the window size and reflectance of the surfaces in rows. As in the calculation above, the identifier for each room, defined in Table 1, is on the upper-right side of each section. As in the previous calculation, the average power consumption of lighting smart controls are found in Table 8, which shows the electric consumption in W/m 2 for each Case study.  As in the previous calculation, the average power consumption of lighting smart controls are found in Table 8, which shows the electric consumption in W/m 2 for each Case study. As seen in Table 8 and deduced from the previous trial, the deep rooms need higher power consumption for electric lighting, since daylight cannot reach the back of the room. In this case, window size is decisive in reducing the power consumption, given that large windows consume between 20% and 50% less power than small windows.

Quantification of Power Consumption in London
Comparison of Tables 6 and 8 shows that the room model in the London location requires approximately 13% more energy than the Stockholm location. The difference between the locations only tends to converge for deep rooms, given that the daylight is not sufficient to light the entire venue in either case. It can therefore be deduced that weather conditions are more significant than latitude in the calculation of power consumption in electric lighting.
As in the previous case, smart controls notably reduce power consumption in electric lighting. The dimming controls save nearly 30% of energy compared to the typical On/Off controls, while the dimming controls with separate rows reduce power consumption by up to 50%.
Following on from Figure 5 and Table 8, Table 9 shows the annual energy saving, based on proposed room dimensions, window size, surface reflectances and smart controls. As before, the annual energy saving is calculated comparing the average power consumption of the smart controls to that of the conventional On/Off control.   As seen in Table 9, converging with the results shown for the Stockholm location, the energy saving is higher for rooms with small windows, except in the case of deep rooms. As above, it can be concluded that a room with a small window can compensate for power consumption using a dimming control with two separate rows (Case study 3), compared to other rooms with larger windows and less efficient smart controls (case studies 1 and 2).
As deduced from Figure 5 and Table 6 above, owing to worse weather conditions the power consumption for London is higher than that for Stockholm. However, comparison of Tables 7 and 9 shows that the energy saving in London is higher for narrow rooms. In fact, rooms 3.00 m deep increase energy saving by almost 20% for dimming controls compared to Stockholm. The opposite occurs in the case of deep rooms, as the energy savings of dimming controls are slightly higher in Stockholm.

Quantification of Power Consumption in Madrid
As above, Figure 6 describes the cross sections for all room models for the Madrid location, defining the dynamic daylight metrics and the average power consumption at the central axis. This figure follows the same structure as the previous ones. The identifier for each room, defined in Table 1, is on the upper-right side of each section. Table 10 shows the average power consumption of lighting smart controls in W/m 2 , based on the case studies.
In line with previous results Table 10 shows that the power consumption in electric lighting for deep rooms with conventional On/Off controls is similar for all locations, irrespective of weather conditions and the reflectance of the inner surfaces. This is because daylight cannot reach the illuminance threshold in the entire room. Therefore, the use of smart controls is even more advantageous in locations with clear skies.
Following the analysis and results of Table 10, except for deep rooms with a low reflectance, large windows (window-to-façade ratio of 90%) consume between 40% and 75% less in lighting than small windows (window-to-façade ratio of 30%). It can be deduced that the impact of a large window on energy efficiency is higher for sites with better weather conditions, and a higher sky luminance.  Figure 6. Dynamic daylight performance metrics and average power consumption according to different smart controls for room models located in Madrid.  From the above, the room model in the Madrid location consumes almost 40% less power consumption than London and almost 35% less than Stockholm.
As seen above, the smart controls are decisive in controlling power consumption in electric lighting. Extending this statement to all the locations studied, the dimming controls (Case study 2) produce an energy saving of close to 30% compared to the typical On/Off controls, while the dimming controls with separate rows (Case study 3) reduce power consumption by up to 55%. Moreover, the On/Off lighting control with separate rows (Case study 1) can reduce power consumption by up to 20% compared to the On/Off system with one row for all luminaires.
In accordance with the results of Figure 6 and Table 10, the annual energy saving is determined in Table 11, based on the proposed room dimensions, window size, surface reflectances and smart controls. As above, the annual energy saving is calculated by comparing the average power consumption of the smart controls proposed to that produced by the conventional On/Off control.
As seen in Table 11 and previously, energy saving is higher for rooms with small windows, due to the high dependence on electric lighting. Deep rooms are an exception, as the conventional On/Off control is almost always on, regardless of window size. As in the cases above, a room with a small window can save more energy than other rooms with larger windows and less efficient smart controls by using dimming control with two separate rows.  Table 13 shows that adequate NPV10 is obtained in the same way for rooms 6.00 m deep or more (average NPV10 of 497.90 € compared to −44.83 € in the case of 3.00 m deep), and especially for rooms 8.00 m wide or more (average NPV10 of 668.22 €). However, unlike the payback period, the NPV10 shows that the more complex the control system in larger rooms, the greater the economic benefits obtained after 10 years. Using a smart control instead of an On/Off manual control with two separate rows (case 1) ultimately provides an economic saving of 73% in the case of dimming control for all luminaires (case 2), and 126% for smart control with two separate rows (case 3), as energy savings are greater in the long term despite their higher initial investment costs.
As with power consumption, the profitability of the smart systems is higher for rooms at least 8.00 m wide and small windows, except in the case of deep rooms with dark surfaces. For example, all NPV10 values of Case study 3 with small windows (window-to-façade ratio of 30%) are equal to or greater than the NPV10 values of Case study 1 with the largest windows (window-to-façade ratio of 90%).

Cost Effectiveness in London
The PP for the three control systems for the London room model is shown in Table 14. Considering the sky conditions and as shown in Table 4, in general PP is 33.5% lower for all the London hypotheses since both the annual electric energy cost in 2016 (EC2016) and the annual growth rate (AGRE) in the United Kingdom are higher than in Sweden. Nevertheless, the use of these three system controls is still advantageous to rooms 6.00 m deep or more (PP average value of 3.3 years compared to 10.1 years in the case of 3.00 m deep), except when they are 12.00 m wide or more (PP average value of 6.7 years). As in the case of Stockholm, the wider the deep rooms, the lower the PP for all the control systems (PP average value of 2.5 years in rooms 8.00 m wide or more), resulting in an average PP decrease of 23.8% compared to Stockholm.
As seen in Stockholm, Table 15 shows that adequate NPV10 is related to depths equal to or greater than 6.00 m (average NPV10 of 693.32 €, a 39.3% increase compared to Stockholm), especially for widths of 8.00 m or more (average NPV10 of 904.60 €, an increase of 35.4%). In the same way, the most complex control systems save most after 10 years, as the average economic saving for smart control systems compared to case 1 is 88% for case 2 and 156% for case 3. These economic results are more significant than the Stockholm ones (a NPV10 increase of 24.3%, 34.8% and 40.7% in the three cases) due both to the higher electric energy cost in the United Kingdom and the higher annual energy saving of lighting control systems in London.  35.6% compared to Stockholm and London, respectively), obtaining the highest economic savings for rooms 8.00 m wide or more (average NPV10 of 1209.58 €, an increase of 81.0% and 33.7%, respectively), especially for rooms 9.00 m deep. In the same way, the greatest economic savings after 10 years are obtained when smart systems are used, with an average NPV10 improvement compared to case 1, of 74% for case 2 and 123% for case 3. Thus, given that the Spanish electric energy cost forecast is the highest of the three locations under study, and that Madrid is the location with the highest annual energy saving in the deepest rooms, the economic saving increases using the systems from cases 2 and 3 in these wide, deep rooms was 82.6% and 79.4% compared to Stockholm, and 35.4% and 27.5% compared to London, respectively. As shown in the case of Stockholm and London, using smart systems in rooms at least 8.00 m wide with small windows is more profitable than installing manual On/Off control systems with two separate rows, except in the case of deep rooms with dark surfaces.
According to the results for the three locations above, it can be concluded that for rooms at least 6.00 m deep and 8.00 m wide, the more complex the control system, the greater the profitability obtained after 10 years. Moreover, due to the higher initial investment costs of the smart control systems, the simpler the control system, the sooner the cost is amortized.