INCORPORATING SKY LUMINANCE DATA MEASURED BY EKO SCANNER WITH A SCANNING SKY SIMULATOR FOR PREDICTING DAYLIGHT QUANTITY IN BUILDINGS Jianxin Hu, PhD. College of Design North Carolina State University Brooks Hall, Box 7701 Raleigh, NC 27695 e-mail: jhu3@ncsu.edu Wane Place, Ph.D. College of Design North Carolina State University Brooks Hall, Box 7701 Raleigh, NC 27695 e-mail: wayne_place@ncsu.edu Christoph Konradi College of Design North Carolina State University Brooks Hall, Box 7701 Raleigh, NC 27695 e-mail: ckonrad@ncsu.edu ABSTRACT A 1/6-dome Scanning Sky Simulator is evaluated in this study as a Climate-Based Daylight Modeling tool for predicting daylight quantity in buildings on an annual basis. This method involves establishing sky luminance distribution as climate data input, testing physical models under the Scanning Sky Simulator as the simulation procedure, and calculating interior daylight quantity based on the concept of Daylight Coefficient as the underlying theory. As a physical simulation tool, the Scanning Sky Simulator is assessed in this study by comparing the performance data acquired from testing two typical daylighting designs under the device with the data collected from monitoring the same designs under the real sky. The results show that the differences between the predicted daylight quantities by the simulator and the measured quantities under the real sky are within 10% in average for the two daylighting systems tested. Key Words: Daylighting, Scanning Sky Simulator, Sky Scanner, Climate-based Daylight Modelling, Light Quantity 1. INTRODUCTION The traditional daylighting research approach the Daylight Factor method only addresses overcast conditions and leaves out important design factors, such as building orientation. Consequently, it would be more practical to assess daylighting systems in areas with a highly luminous climate by a climate-based method, in which case various types of sky conditions in the course of a year are all taken into considerations [1]. Although computer-based simulation packages, such as DAYSIM, are increasingly used for Climate-Based Daylight Modeling (CBDM) [2 & 3], conducting physical experiments is still crucial for daylighting research, education, and application, for the following reasons: - Manipulation of real objects (the models) engages the designers and clients more readily than screen or paper representations. - Light quality and spatial perception can be difficult to assess by computer simulations. A large enough scale model allows people to assess light quality by observing through the view ports provided on the walls, so that the observer can be immersed in the luminous surround. - There are still many materials, devices and products that cannot be easily characterized by existing software tools. - Fast turnaround: results can be immediate, and although a model needs to be constructed, they can often be very simple and still provide good results. However, assessing annual daylighting performance by monitoring the physical scale-models of the design options in a full year period is not feasible in practice. Simplified methods are thus needed to conduct these assessments in a timely manner. This project is intended to develop such a simplified method. It involves the following three steps: Step 1: Using historical meteorological data that are available at the project location to establish sky luminance
distribution during the course of a year. This step can be achieved by directly measuring sky luminance using a sky scanner (e.g. EKO scanner) or by deriving sky luminance from local TMY weather data (e.g. Horizontal, global, and direct normal solar irradiances). For the purpose of this study, the former approach is adopted. Step 2: Testing physical models under an artificial sky to quickly establish the correlation between the interior illuminance at the points of interest and the sky conditions with different characteristics. Various types of artificial sky have been developed in the literature. Two examples are mirror-box artificial sky and Scanning Sky Simulator. Although the former can effectively simulate an overcast sky, the latter would be more appropriate for climate-based daylighting research, as it can simulate various types of sky conditions. A detailed description of this tool will be provided in a later section of this paper. Step 3: Incorporating the sky luminance data obtained in step 1 with the correlations established in step 2 to predict daylight quantity in buildings. This last step is essentially a calculation procedure based on the concept of Daylight Coefficient [4]. As outlined in the above three steps, the intent of this paper is to report the procedures and the underlying theory of using a Scanning Sky Simulator, coupled with the sky luminance data collected by an EKO sky scanner, to predict daylight quantity at a given point in buildings on an annual basis. The validation process and the accuracy of this method are also reported. Developed by Dr. L. Michel, etc., a similar Scanning Sky Simulator was established in EPFL Solar Energy and Building Physics Laboratory [5]. Multiple studies have been done by the research group focusing on using the simulator for daylight quantity predictions. In an earlier paper, Michel, etc. reported a study on using the device based on the Daylight Factor approach [6]. In a later paper, a Partial Daylight Factor Method was introduced by employing the same device [7]. Compared to these previous studies, the method proposed in the present research features the following uniqueness: - It incorporates sky luminance data collected by a sky scanner as climate data input - It is based on the concept of Daylight Coefficient instead of Daylight Factor. - It is intended to assess climate-based daylight performance, which is on a year-round basis. 2. METHODS 2.1 Tregenza s Sky Model: Basis of Defining the Sky To define a particular sky condition, Commission International De L Eclairage (CIE) recommended using - segment equal-area subdivision based on 8 equal altitude bands [8]. This method, developed from Tregenza s early work [9], allows each zone to be considered as a point source without noticeable error (Figure 1). Figure 1: Sky luminance distribution defined by patches 2.2 Sky Luminances: Daylight Climate Data Input An IDMP (International Daylighting Measurement Program) Research Grade station (complies with CIE108-1994) has been established at rooftop of the Daylighting Research Lab located on NC State University campus to collect daylight climate data for a full-year period in Raleigh, North Carolina. In addition to meteorological data, such as solar irradiances, daylight illuminances, air temperature and relative humidity, sky luminance and radiance distributions are also collected by an EKO sky scanner (Figure 2). The way that the sky scanner divides up the sky into small segments is consistent with the CIE standard of sub-dividing the sky dome. The sensor head rotates in altitude and azimuth to measure the luminance at circular sky patches by scanning the sky hemisphere. Each scanning cycle takes about 4.5 minutes and measurements are recorded every 10 minutes. Figure 2: EKO Scanner collecting sky luminance
2.3 Scanning Sky Simulator: The Simulation Tool A Scanning Sky Simulator was constructed as an experiment-based CBDM tools. The device is designed to represent one-sixth of the sky dome, which is also based on Tregenza's model of sky patches. Starting with a sixth of a hemisphere including 25 light sources, the hemisphere can be rebuilt by a six-step scan (Figure 3). In order to validate the proposed method, the physical models tested under the Scanning Sky Simulator are the exact copies of the models being monitored under the real sky on an annual basis as reported in Section 2.4. Figure 4: Scanning Sky Simulator at NC State Figure 3: 1/6 of sky dome including 25 patches Each of the light sources was made up of four LED lights (Flood light with 550 Lumens / 10 watt each). The radius of the simulator is 6 feet and 6 inches. The surrounding environment is darkened when experiment is being conducted. The device also includes a turning table at the center, which sets the relative angles between the physical model and the sky simulator (Figure 4). LI-COR light sensors are installed inside the model for measuring illuminance values at the points of interest on the task plane. A data logger (Campbell Scientific CR1000) mounted on the turning table collects and stores the data measured by the light sensors. During the first of the 6-step scan, the 25 light sources are turned on one at a time, which means, at any given point of time, only one light source is on. Simultaneously, inside the physical model, the illuminance values contributed by this particular light source are recorded. This procedure is performed for 25 times for all 25 light sources, which completes the first of the 6-step scan. Then the turning table rotates for 60 degrees for the second step. When the entire 6-step scan is completed, the contributions of each of the sky patches to the interior illuminances become available. 2.4 Monitored Physical Scale Models: The Validation Tool Multiple scale models (1/2 = 1-0 ) of typical daylighting systems have been constructed and monitored in the close proximity to the climate data collection station (Figure 5). Two of these models are used for validation purposes. Specifically, these models are: south-facing apertures with overhang and lightshelf and north-facing saw-tooth apertures. The models are grouped in a cluster and are protected from weather by a clear acrylic dome with a visible light transmittance of 95%. Multiple photocell sensors are placed in each model for measuring light quantities on a year-round basis. As mentioned earlier, the exact copies of these two models are reconstructed and tested in the Scanning Sky Simulator. By comparing illuminance values acquired from testing these models under the Scanning Sky Simulator with the data collected from the same models monitored under the real sky, the effectiveness of the sky simulator and the calculation algorithms can be validated. Since the lighting output of all the light sources is a constant and a given, Daylight Coefficient can be calculated between a sky patch and the illuminance value at a point of interest inside the model. The definition of Daylight Coefficient and the calculation procedure will be discussed in detail in Section 2.5. Figure 5: Scale models being monitored
2.5 Daylighting Coefficient: The Underlying Theory The Definition of Daylight Coefficient is illustrated in Figure 6. α Sky element number (1 through ) Under the Scanning Sky Simulator, a Daylight Coefficient can be established in the same fashion: dc! =!e! "!s! [2] Figure 6: Definition of Daylight Coefficient - Image source: DAYSIM User Manual [4] Assume the luminance of a sky element is and its angular size is!s!. This sky element produces an illuminance of at a point in a room. Under a real sky, Daylight Coefficient is defined as [4]: where:!e! DC! =!E! "!S! [1] DC! Daylight Coefficient of sky element α at a point under real sky!e! Illuminance produced at the point by sky element α. This value cannot be physically measured under a real sky, because it is impossible to separate sky element α from the rest of the sky. However, the sum of!e! from all sky elements can be measured by an illuminance sensor placed inside the physical model under a real sky Luminance of sky element α. This value is measured by the EKO sky scanner.!s! Angular size of sky element α. It is assumed that this value is approximately 2π/ (the solid angle of a hemisphere is 2π).! "S! This is in fact the direct normal illuminance on an unobstructed plane facing a sky element where: dc! Daylight Coefficient of sky element α at a point under the sky simulator!e! Illuminance produced at the point by sky element α. This value can be measured by an illuminance sensor inside a physical model when only the light source at sky element α is turned on. Luminance of sky element α. This value is determined by the lighting output of the light sources. Since all light sources are established by the same type of LED lamps, is assumed to be the same for all sky elements in the sky simulator.!s! Angular size of sky element α. This value is also approximately 2π/.! "s! Similarly this is the direct normal illuminance on an unobstructed plane facing a sky element, which is a light source in the sky simulator. This illuminance value can be measured. α Sky element number (1 through ) Daylight Coefficient solely depends on the nature of the architectural space and it s surrounding physical environment. It is independent of sky luminance ( ), because if sky luminance changes, the illuminance (!E! ) will change in the same proportion. Therefore for a particular point in an architectural space, the Daylight Coefficients developed under the real sky and under the sky simulator shall be the same theoretically: DC! = dc! [3] Therefore:!E! =!e! [4] "!S! "!s! As mentioned above, the term! "s! is in fact the direct normal illuminance on unobstructed plane facing a sky element and it should be a constant for a given sky simulator, since all light sources are configured the same way. e patch is
then used to represent this term (it is measured at 800 lux for the sky simulator constructed in this study). Then Equation [4] becomes: or!e! "!S! =!e! e patch [5]! "S!! "e! e patch = "E! [6] For elements combined:!s! # "!e! = #!E! [7] e patch!=1!=1 Since!S! = 2π/, equation [7] becomes: The sum of 2! # L "! "e " = #"E " [8]! e patch "=1 "=1 contributed by all sky elements: "!E!, is the measured illuminance at the point of!=1 interest, and can be measured under the real sky by using an illuminance sensor placed inside a physical model. If E is used to represent this sum, Equation [8] becomes: 2! # L "! "e " = E [9]! e patch 3. VALIDATION "=1 3.1 Data Collection for Validation Purposes To validate Equation [9], the following data have been made available: E!E! An EKO sky scanner collects sky luminance distribution data in Raleigh, NC. Each scan provides luminance values for all sky elements defined by the Tregenza model Physical scale models of typical daylighting designs are observed simultaneously with the sky luminance collection at the same location in Raleigh. Multiple illuminance sensors are placed in each model to collect actual illuminance data at the same time interval as the sky scanner.!e! While the above-mentioned physical models are being monitored at the outdoor station, duplicates of these models including the interior illuminance sensors are reconstructed and tested under the sky simulator. At any given time, only one sky element of the device is turned on and the corresponding interior illuminance (!e! ) contributed by this element is measured. A total of!e! are recorded for each scan. e patch This is the direct normal illuminance at the center of the simulator facing a particular sky element. Since all light sources are the same in the simulator, this is a constant for all sky elements. It is measured at 800 lux. 3.2 Validation Procedure Hourly datasets (3,213 data points) including all above variables are input into equation [9]. To account for any errors, such as the ones caused by equipment calibration, a correction factor K is added to equation [9]: 2! K # L "! "e " = E [10]! e patch "=1 Theoretically, K is expected to remain 1 for all data sets. Practically one can predict that K fluctuates around 1. However, the results have shown that the fluctuation correlates significantly to the clearness of the sky. An overcast sky tends to result in a lower K value than 1, whereas a clearer sky generates a higher K value than 1. While the fluctuation of K factor partially results from the random errors of the equipments, it is a challenge to explain the strong correlation between K and the clearness of the sky, because presumably the sky condition has been quantified and characterized by the sky luminance data collected by the sky scanner. Ineichen found in an earlier study that the EKO and the PRC Krochmann scanners produced larger discrepancies under clear skies, which implies that the EKO scanner can make significant systematic errors that are dependent upon the clearness of the sky [10]. To address this issue caused by the nature of the EKO scanner, the entire data points are then divided into 8 bins based on sky clearness (ε) [11]. A separate K factor is derived for each bin. Based on the two daylighting models tested, the K values are derived as follows (Figure 7):
Correction Factor K 1.80 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 Sky Clearness (ε) Figure 7: Correction Factor (K) developed from Equation [10] Model 1 is the saw tooth apertures facing north. Model 2 is the light shelf system with an overhang facing south. The K values developed from these two models are fairly close. The averaged Ks of these two curves are (Table 1): Table 1: Correction Factor (K) Model 1 Model 2 1 2 3 4 5 6 7 8 threshold is met by daylit alone [12]. This threshold is essentially the design illuminance level for a particular space (e.g. 300 Lux for offices). Recently proposed by Rogers, Continuous Daylight Autonomy is a metric that considers partial credit when the daylight illuminance lies below the minimum illuminance level. For example, in the case where 300 lux are required and 150 lux are provided by daylight at a given time step, a partial credit of 150lux/300lux=0.5 is given for that time step. An upper threshold, which is usually equal to ten times the design illuminance, is also established to address overlit (glare) issues. In the example above, the upper threshold would be 300lux X 10 = 3000 lux. If the daylight level falls between 300lux and 3000lux, the cda is 1. For the same sensor location in that model, another set of UDI and DAcont can also be developed from the measured illuminances under the real sky (Section 2.4). The predicted and measured UDIs and cdas are then compared for each sensor location in both Model 1 and Model 2. The results are shown in Figure 8 through Figure 11. Useful Daylight Index 1.2 1 0.8 0.6 3.3 Prediction by Using the Validated Algorithm Once K factors are developed for Equation [10], the sky simulator, coupled with sky luminance data collected by the sky scanner, can be used to predict daylight quantities ( E ) in buildings. The process would involve building a physical scale model and, at any given point in the model, measuring the illuminance (!e! ) level contributed by each sky element of the device. For each sensor location in a particular model, two types of climate-based daylighting performance indicators, Useful Daylight Index (UDI) and Continuous Daylight Autonomy (cda), are developed respectively from the daylight quantities predicted by the procedures above. 0.4 0.2 0 1 2 3 4 5 6 7 8 8 Sensor Locations inside Model 1 Figure 8: Measured and predicted UDIs for 8 sensor locations in Model 1 Saw tooth apertures facing north Useful Daylight Index 1.2 1 0.8 0.6 Measured Predicted UDI is defined as the annual occurrence of illuminances at any given point on the work plane that are within a range considered useful by occupants. The useful range was identified based on a survey of previous studies on occupant perceptions and preferences. The widely accepted rang is 100 Lux to 2000 Lux [12]. 0.4 0.2 0 Measured Predicted 1 2 3 4 5 6 6 Sensor Locations inside Model 2 The earlier definition of Daylight Autonomy is the percentage of the year when a minimum illuminance Figure 9: Measured and predicted UDIs for 6 sensor locations in Model 2 Light shelf with overhang facing south
Continuous Daylight Autonomy 1.2 1 0.8 0.6 0.4 0.2 0 1 2 3 4 5 6 7 8 8 Sensor Locations inside Model 1 Figure 10: Measured and predicted cdas for 8 sensor locations in Model 1 Saw tooth skylight facing north Continuous Daylight Autonomy 1 0.8 0.6 0.4 0.2 0 1 2 3 4 5 6 6 Sensor Locations inside Model 2 Figure 11: Measured and predicted cdas for 6 sensor locations in Model 2 Light shelf with overhang facing south 4. Conclusion Measured Predicted Measured Predicted The results show that the differences between the predicted daylight quantities by the Scanning Sky Simulator and the measured quantities under the real sky are within 10% in average for the two daylighting systems tested. As mentioned earlier, assessing annual light quantity by physical testing can be difficult, since monitoring a model of a design option on a year-round basis is extremely time consuming. However, as proven by the above results, the simplified experimental method reported in this paper, including the simulating procedure and the algorithms, can be reasonably accurate for predicting daylight levels in building. The method is also efficient. Once the physical models are constructed, the testing process under the simulator can be completed in a timely manner. The six-step scan typically takes approximately 30 minutes. 5. Limitations and Future Research Only two exemplary models were tested to validate the device and the algorithms. More models will be tested as the next step to strengthen the validation. It is suspected that the limitation of the EKO sky scanner leads to the fact that the correction factor (K) factor correlates to sky clearness. The scanner samples the sky in a circular view angle of 11 degrees for each sky patch (Figure 1). Under sunny sky conditions, if the solar positions happen to fall in between the circles, sky luminance readings are expected to be inaccurate. Instead of using a sky scanner, an alternative would be to derive sky luminance distribution from solar irradiance data, which are widely available, by using Perez s sky luminance model [11]. 6. REFERENCES [1] Mardaljevic, J., Examples of Climate-Based Daylight Modelling, CIBSE National Conference 2006: Engineering the Future. [2] Reinhart C F, Walkenhorst O, Dynamic RADIANCEbased daylight simulations for a full-scale test office with outer venetian blinds. Energy & Buildings, 33:7 pp. 683-697, 2001. [3] Reinhart C F, Herkel S, The simulation of annual daylight illuminance distributions a state-of-the-art comparison of six Radiance-based methods. Energy & Buildings, 32 pp. 167-187, 2000. [4] Reinhart C F, Tutorial on the Use of Daysim Simulations for Sustainable Design. [5] http://leso.epfl.ch/skanningsky [6] Michel, L.; Scartezzini, J.-L., Implementing the Partial Daylight Factor Method under a Scanning Sky Simulator, Solar Energy, Elsevier Science Ltd, vol. 72, num. 6 (2002), p. 473-492 [7] Michel L., Roecker C., Scartezzini, J.-L., Performance of a new sky scanning simulator, Lighting Research and Technology, vol. 27, num. 4 (1995), p. 197-208 [8] CIE - Commission Internationale de l'eclairage, Guide to Recommended Practice of Daylight Measurement, CIE Vienna, 1989 [9] Tregenza, P.R., Subdivision of the sky hemisphere for luminance measurements. Lighting Research and Technology 19 (1), 13 14., 1987 [10] Ineichen, P. & B. Molineaux, Characterisation and Comparison of two Sky Scanners : PRC Krochmann & EKO Instruments, IEA Task XVII expert meeting, Geneva, August, 1993 [11] Perez, R., R. Seals & J. Michalsky. All-Weather Model for Sky Luminance Distribution Preliminary Configuration and Validation. Solar Energy, Vol. 50, No. 3 pp. 235-245, 1993. [12] Reinhart C F, Mardaljevic, J., & Z. Rogers, Dynamic daylight performance metrics for sustainable building design, NRCC-48669, 2006