Estimation of Illuminance/Luminance Influence Factor in Intelligent Lighting System Using Operation Log Data

Transcription

1 Estimation of Illuminance/Luminance Influence Factor in Intelligent Lighting System Using Operation Log Data Yuki Sakakibara, Mitsunori Miki 1, Hisanori Ikegami,Hiroto Aida 1 1 Graduate School of Science and Engineering, Doshisha University, Kyoto, Japan Department of Science and Engineering, Doshisha University, Kyoto, Japan Abstract We have proposed the intelligent lighting system which provides the required illuminance to user. We introduce this system to a real office. In office, we measure illuminance/luminance Influence factor, and use this value to control the intelligent lighting system. Illuminance/luminance Influence factor needs update because this value changes office environment, e.g. deterioration of lighting fixture, installation of partition. In order to measure this value, we need to turn on and off the lighting fixture one by one. But we can t measure illuminance/luminance influence factor because it might be uncomfortable to office worker. In this paper, we propose method to update illuminance/luminance influence factor using log data of intelligent lighting system without influence measurement. As a result of verification, it was confirmed that this propose method is effective to update Illuminance/luminance influence factor. Keywords: Lighting Control, optimization, Saving energy 1. Introduction In recent years, research has been widely conducted on the impact of the office environment on the productivity of employees, and it has been reported that improving an office environment will lead to improvements in the intellectual productivity of employees[1][][]. In particular, according to research that focuses on the lighting environment as one of the aspects of an office environment, providing each employee with the optimum amount of brightness (illuminance) for work is an effective means of improving the office environment[]. Moreover, the amount of energy consumed in office buildings is increasing year by year, and this increase in energy consumption has become a problem. The power consumed by lighting in office buildings accounts for approximately % of the total amount of power consumed[6], and thus reducing the power consumed by lighting will lead to a reduction in the amount of power consumed by an office. With that in mind, the authors are conducting research and development on a lighting system (hereafter, Intelligent Lighting System)[] that realizes individual illuminance for the purpose of improving the intellectual productivity of employees in an office environment and conserving the amount of energy consumed by offices [9]. The Intelligent Lighting System was tested and verified in a laboratory, after which a prototype system was introduced in multiple offices where demonstration testing was conducted for the purpose of achieving practical applications, and the usefulness thereof was verified. The Intelligent Lighting System is configured by connecting multiple light controllable light fixtures with built-in microprocessors, multiple illuminance meters, and power sensors to a network. By using the optimum technique based on information obtained from the illuminance meters and the power sensors, each light achieves a lighting pattern that both satisfies the target illuminance set by each user and reduces power consumption. With the Intelligent Lighting System, the impact that the luminance of each lighting unit has on the illuminance of each meter (hereafter, illuminance/luminance influence factor) is used to implement lighting control. The Intelligent Lighting System introduced in actual offices measures the illuminance/luminance influence at the time of system introduction (hereafter, illuminance/luminance influence factor measurement method), and the value thereof is used to control the lighting. This method measures the actual illuminance/luminance influence factor that each light has on each illuminance meter by turning off all the lights in the targeted environment and then turning on one light at a time.. Intelligent Lighting System.1 Overview of an Intelligent Lighting System The intelligent lighting system, as indicated in Figure 1, is composed of lights equipped with microprocessors, illuminance sensors, and electrical power meters, with each element connected via a network. Electric meter Network Lighting Fixture Illuminance sensor Microprocessor Figure 1: The construction of a intelligent lighting system

2 Individual users set the illuminance constraint on the illuminance sensors. At this time, each light repeats autonomous changes in luminance to converge to an optimum ligthing pattern. Also, with the intelligent lighting system, positional information for the lights and illuminance sensors is unnecessary. This is because the lights learn the factor of influence to the illuminance sensors, based on illuminance data sent from illuminance sensors. In this fashion, each user s target illuminance can be provided rapidly. The most significant feature of the intelligent lighting system is that no component exists for integrated control of the whole system; each light is controlled autonomously. For this reason, the system has a high degree of fault tolerance, making it highly reliable even for large-scale offices.. Control of the Intelligent Lighting System In the intelligent lighting system, the algorithm where Simulated Annealing (SA) is improved for lighting control (Adaptive Neighborhood Algorithm using Regression Coefficient: ANA/RC) is used to control luminance intensity for each lighting fixture[][7]. It is possible with ANA/RC to provide the target illuminance with minimum power consumption by making luminance intensity for lighting fixtures the design variable and by using the difference between the current illuminance and target illuminance as well as power consumption as objective functions[7]. Furthermore, by learning the influence of each lighting fixture on each illuminance sensor using the regression analysis and by changing the luminance intensity depending on the results, it is possible to promptly change to the optimal luminance intensity. This algorithm is effective to solve the problem which the objective function is near monomodal function and changes in real time. And it is possible to cut power consumption about % in actual office[6][9].. Illuminance/luminance Influence Factor As explained in the preceding section, the light control algorithm (ANA/RC) determines the next appropriate luminance and endeavors to improve convergence speed by computing the regression coefficient (illuminance/luminance influence factor) from the amount of change in luminance and the amount of change in illuminance[7]. Therefore, understanding the accurate value for the illuminance/luminance influence factor is important for the control of each light by the Intelligent Lighting System. However, because the luminance of each light is randomly changed, the possibility of correlation temporarily occurring with the change in the luminance of each light increases as the number of lights is increased. When this occurs, a phenomenon is observed in the Intelligent Lighting System introduced in the actual offices wherein the correlation between the change in luminance of fundamentally irrelevant lights and the change in illuminance measured with an illuminance meter temporarily increases, and an erroneous illuminance/luminance influence factor is obtained. At this time, phenomena such as a temporary brightening of unnecessary lights occurs. This type of phenomenon is temporary and has almost no impact on the illuminance required by each employee or on overall power consumption. However, when the Intelligent Lighting System was introduced in actual offices, a questionnaire given to the employees therein indicated that the employees were slightly concerned about the phenomenon. In order to improve this, in offices with no movement of the illuminance meter, we proposed a technique to provide the users with more stable lighting control by measuring the actual illuminance/luminance influence factor when the Intelligent Lighting System was introduced and storing that value in a database rather than dynamically estimating the illuminance/luminance influence factor. The technique for measuring the actual illuminance/luminance influence factor computes the illuminance/luminance influence factor from the relationship expression shown by Equation () by causing one light at a time to turn on when the Intelligent Lighting System is introduced. This technique is called the illuminance/luminance influence factor measurement method. The illuminance/luminance influence factor is a number that is obtained by combining the perspective [far and near] relationship of the lighting and the illuminance meter, reflection from the wall surfaces, light shielding by partitions, and such, and quantifying the impact of the luminance of a single light on the illuminance of a single meter.. Issues in Actual Offices We are conducting demonstration testing of the Intelligent Lighting System in multiple actual offices in the Tokyo area in Japan. In the actual offices in which the Intelligent Lighting System was introduced, there are many fixed seats and no movement of the illuminance meters, and therefore, the illuminance/luminance influence factor computed using the illuminance/luuminance influence factor measurement method described in the preceding section is used to control the lighting. However, the illuminance/luminance influence factor varies depending on the following types of changes in the lighting environment. When the above-described changes in the lighting environment occur in an actual office, the illuminance/luminance influence factor must be re-measured and updated using the illuminance/luminance influence factor measurement method. However, measuring the illuminance/luminance influence factor in an actual office creates problems for employees trying to work in the office because the lights are repeatedly switched on and off. Therefore, implementing this type of measurement is generally not allowed in many cases. Moreover, implementing measurement tests during the night when employees are not present is also not allowed from

3 the viewpoint of security. Thus, as a technique to estimate the illuminance/luminance influence factor for this research, a technique is proposed that is based on mathematical programming using operation log data for the Intelligent Lighting System.. Estimating the Illuminance/Luminance Influence Factor Using Operation Log Data A technique for estimating the illuminance/luminance influence factor using operation log data for the Intelligent Lighting System is proposed. By using this technique, the illuminance/luminance influence factor can be updated without interrupting the work going on in the office, and the illuminance/luminance influence factor can be estimated in accordance with changes in the lighting environment without incurring any cost or labor to measure the actual illuminance/luminance influence factor. An example of log data used in the estimation technique is shown in the Figure, and as shown therein, the luminance value for the number of lights therein and the illuminance value for the number of illuminance meters are used to estimate the illuminance/luminance influence factor. Time luminance luminance (light 1) (light ) measured_illuminance (sensor A) target_illuminance (sensor A) 9: : : : : : : : : : : : : : : : Figure : Log data of a intelligent lighting system The relational expression of Equation (1) shows the relationship between the illuminance value obtained from each illuminance meter and the luminance value for the multiple lights. Therefore, if there is a significant difference between the luminance value for each light in the operation log data and the illuminance value of each illuminance meter, the relational expression of Equation (1) can be used to compute the illuminance/luminance influence factor by setting up a system of equations for the illuminance/luminance influence factor. I i = m R i,j L j (1) j=1 I x : IlluminanceL y : Luminance, R i,j Illuminance/Luminance influence factor m : number of Lighting fixture However, in some cases, the luminance value in the operation log data and the illuminance value do not vary. In addition, a margin of error exists also in the observed illuminance and luminance values. From these points, the illuminance/luminance influence factor is computed by solving an optimization problem to minimize the margin of error rather by than solving a system of equations. The optimization problem finds the objective function expressed by the square of the difference of the estimated illuminance and measured illuminance for each data series when the luminance of the lighting and the illuminance of the meters at the same time in the operation log data are used as a single data series. The summation for the data series of the objective function thereof is minimized as a single objective function, and the illuminance/luminance influence factor is estimated. The objective function is shown in Equation (). min : F = f i (R j,k ) = E i,j = d f i (R j,k ) () i=1 n (E i,j I i,j ) j=1 m R j,k L i,k k=1 d number of Datam : number of Lighting fixtures, n : number of sensorsl : luminance intensity in an operation log data I illuminance intensity in an operation log datae : estimate illuminancer j,k : illuminance/luminance influence factor(design value) The optimization problem for the illuminance/luminance influence factor estimation shown in Equation () is nonlinear, and therefore the method of steepest descent is used here as the mathematical programming technique.. Testing Overview In this chapter,i examine the usefulness of the proposed method by carried out the following experiment. Determine if estimation of the illuminance/luminance influence factor is possible using the proposed technique. (Accuracy Verification Test) Determine if the illuminance/luminance influence factor can be updated in accordance with changes to the lighting environment. (Lighting Environment Variation Test) In the accuracy verification test, the illuminance (true value) computed from the actual illuminance/luminance influence factor measured using the illuminance/luminance

4 influence factor measurement method and the illuminance/luminance influence factor estimated by the proposed technique were used to compute the illuminance (estimated value), and with that illuminance (estimated value) as the evaluation target, the usefulness of the proposed technique was verified. In the lighting environment variation test, partitions were installed in order to simulate changes in the lighting environment while the Intelligent Lighting System was in operation. When partitions are installed, the illuminance/luminance influence factor of the luminance of the lights on the illuminance of the meters changes, and as a result, the lighting pattern of the lights worsens. In this case, we verified whether or not an optimum lighting pattern can be realized by using the log data after installation of the partitions to update the illuminance/luminance influence factor..1 Testing Enviroment A plan view of the model environment used in the verification testing is shown in the Figure. In this testing, three illuminance meters and 1 white fluorescent lamps were used. In addition, work by employees in an actual environment was simulated, the illuminance meters were installed.7m above the floor at positions with 1.9m from the illuminance meters to the ceiling, and testing was conducted. For the light fixtures, white fluorescent, light controllable lamps (FHPEN) from Panasonic with luminance ranging from a minimum lighting luminance of (% lighted: cd) to a maximum lighting luminance of (1% lighted: 1 cd) were used, and for the illuminance sensors, general ANA-F11 type Class A digital illuminance meters were used. Estimation of the illuminance/luminance influence factor in an actual environment was simulated, and the illuminance/luminance influence factor was estimated using 6 minutes of log data in which the employees changed the target illuminance as shown in the Table 1, or more specifically, using 18 steps of operation log data with seconds per step. The history of illuminance used in the estimation is also shown in the Figure. Table 1: The history of target illuminance time [min] sensor A [lx] sensor B[lx] sensor C [lx] Test Results and Considerations(Accuracy Verification Test) The illuminance/luminance influence factor was estimated using the proposed technique. In order to clarify the test results here, measurements equal to or greater than. Lighting fixture A Index of illuminance sensor B Illuminance sensor Figure : Experiment enviroment(acurracy vertification of experiment) Illuminance [lx] sensor B sensor A sensor C C 1 6 Time [min] Figure : The history of measured illuminance lx/cd for the illuminance/luminance influence factor with respect to each illuminance meter are shown in the figure. Moreover, estimations were achieved with the same accuracy for illuminance/luminance influence factors of less than. lx/cd. As also described in section., the illuminance/luminance influence factor is a quantification of the amount of impact that each light has on each illuminance meter. Therefore, the extent of change that actually occurs in the illuminance when each light is lighted, must be verified. With that in mind, next we used the illuminance/luminance influence factor estimated by the technique proposed in this research to evaluate the proposed technique by computing the illuminance margin of error. The illuminance margin of error is the error between the illuminance (true value) computed by multiplying the actual measured illuminance/luminance influence factor by the luminance and the illuminance (estimated value) calculated by multiplying the luminance by the illuminance/luminance influence factor estimated by the proposed technique. The proposed technique was verified by preparing a distribution of the illuminance margin of error using a histogram. In order to verify the illuminance margin of error through

5 Illuminance/Luminance Influence Factor [lx/cd] Measured value Estimated value 1 (a) Illuminance sensor A Freequency [%] 6 Average = 1. lx Max = 8 lx Difference [lx] (a) Illuminance sensor A Illuminance/Luminance Influence Factor [lx/cd] Measured value Estimated value (b) Illuminance sensor B Freequency [%] 6 1 Average = 1.8 lx Max = lx 1 6 Difference [lx] (b) Illuminance sensor B Illuminance/Luminance Influence Factor [lx/cd] Measured value Estimated value (c) Illuminance sensor C Freequency [%] 6 1 Average = 16.8 lx Max = 6 lx 1 6 Difference [lx] (c) Illuminance sensor C Figure : Estimated value of Influence Coefficient and measured value of Influence Coefficient various lighting patterns, the lighting luminance of the lights in the test environment was randomly prepared, and the data thereof was used for the luminance that was used to compute the histogram. The figure shows a histogram of the illuminance margin of error. From the figure, the average illuminance margin of error is below lx. This illuminance margin of error is a level that is not perceivable by people working in an office[8], and one could thus argue that the proposed technique is useful in an actual environment. Moreover, the occurrence of the maximum illuminance margin of error of 8 lx could be attributed to the following causes. Error due to the calibration curve with regards to the luminance and the signal value. Change in the illuminance value due to the temperature of the fluorescent lamps. First, consideration is given to a calibration curve error. The Intelligent Lighting System controls the design variable as luminance. However, in order to cause a light to turn Figure 6: Difference of illuminance on with the luminance thereof, a pulse changes the duty ratio in 6 steps based on Pulse Width Modulation (PWM) and transmits a signal value. In the calculation of the signal value from the luminance value, conversion is based on the relationship (calibration curve) between the luminance measured in advance and the signal value. The calibration curve measures the relationship between luminance and the signal value using an illuminance meter, but with typical illuminance meters, a margin of error exists in the measured values, and the illuminance meter (ANA-F11) used in the testing also generated a margin of error of %. Therefore, because the calibration curve also contains the margin of error of the illuminance meter, it is thought that this is a factor that caused the illuminance margin of error to occur. Next, consideration is given to the characteristics of the fluorescent lamps. It was found that with the fluorescent lamps used in this test, a maximum margin of error of 16 cd was generated in the lighting luminance of the lights due to temperature. Moreover, it is conceivable that a margin

6 of error was generated in the luminance of the lights due to individual differences in the light fixtures as well and to the state of deterioration of each light. It is thought that the margin of error occurred due to these factors.. Test Results and Considerations (Lighting Environment Variation Test) In the lighting environment variation test, a verification was conducted to determine whether or not the lighting pattern of the lights could be improved by appropriately updating the influence factor according to changes in the lighting environment. As shown in the figure, the test environment was the same as the environment used in the accuracy verification test, and to simulate changes in the lighting environment, partitions were installed below the illuminance meters C. A plan view of model environment used in Lighting Environment Variation Test is shoen Figure 7. Lighting fixture A Index of illuminance sensor B Illuminance sensor 1 6 lx 7 % 71 % 9 6 % % 1 1 % lx 6 % % 1 6 lx 1 % 89 % Figure 8: The distribution of luminance(before update illuminance/luminance influence factor) illuminance convergence test as shown in the table was conducted. Also, based on the obtained operation log data, the illuminance/luminance influence factor was updated using the proposed technique. The lighting pattern that realizes the target illuminance using the updated illuminance/luminance influence factor is shown in the Figure 9. Partition Figure 7: Experiment enviromentlighting Environment Variation Test The lighting pattern used to realize the target illuminance for the lights after the partitions were installed is shown in the Figure 8. The size of the circle indicates the strength of the lighted luminance of the lights. Moreover, the target illuminance values of the illuminance meters A, B, and C were lx, lx, and 6 lx respectively. The illuminance/luminance influence factor changes when partitions are installed. As a result, it was found that the light 1 is strongly lighted because the illuminance/luminance influence factor is not updated even though the illuminance/luminance influence factor that the light 1 has on the illuminance meter C decreases. Turning on lights that are not necessary for employees in an office is a factor that has a pejorative effect on energy conservation, and thus this is not an optimum lighting pattern in the Intelligent Lighting System. Next, consideration was given to employees working in an office, and similar to the accuracy verification test, a 1-hour C lx 71 % 76 % 9 6 % % 1 1 % 6 lx 7 % % 1 61 lx 81 % Figure 9: The distribution of luminance(after update illuminance/luminance influence factor) It was found that by updating the illuminance/luminance influence factor in accordance with changes in the lighting environment, the light 1 in the figure dims compared to before the update. Through this, it became possible to improve the lighting distribution of each light. The figure also shows a comparison of power consumption before and after the update. A reduction in power consumption of about W was possible by improving the lighting pattern. From these information, one could argue that the proposed technique is useful in the Intelligent Lighting System.

7 Table : Compare of electoronic power electric power [W] before update 67.9 after update Conclusion This research addressed the issue of not being able to measure the actual illuminance/luminance influence factor in an actual office and proposed a technique for updating the illuminance/luminance influence factor using operation log data for the Intelligent Lighting System. By using this technique, the illuminance/luminance influence factor is updated from operation log data, and as a result, the illuminance/luminance influence factor can be updated without interrupting the work going on in an office. Moreover, the illuminance/luminance influence factor can be estimated in accordance with changes in the lighting environment without incurring any cost or labor to measure the illuminance/luminance influence factor. Thus, light control tests were conducted at this university, and the log data thereby obtained was used to verify the usefulness of the proposed technique. From the test results, it was found that the illuminance/luminance influence factor can be updated in accordance with changes to the lighting environment, and that the illuminance required by employees in an office can be realized with an optimum lighting pattern. Moreover, improving the lighting pattern also enabled a reduction in power consumption. We believe that by updating the illuminance/luminance influence factor using the proposed technique for the Intelligent Lighting System that was installed in actual offices, a more optimal lighting pattern can be realized for employees working in offices, and that this will contribute significantly to the widespread use of the Intelligent Lighting System. [8] T.Shikakura, H.Morikawa, and Y.NakamuraResearch on the Perception of Lighting Fluctuation in a Luminous Offices Environment Journal of the Illuminating Engineering Institute of Japan, Vol.8, pp.6-11 [9] M. MikiF. KakuT. HiroyasuM. YoshimiS. TanakaJ. Tanisawaand T. NishimotoConstruction of Intelligent Lighting System Providing desired Illuminance Distributions in Actual Office EnviromentJournal of the Institute of Electronics of Japan, Vol.J9-D, pp References [1] N.Nishihara,S.Tanabe,Subjective experiment on productivity under Moderately hot Enviroment, J.Environ. AIJ,No.68, pp.-9, [] M.Miki, and T.Kawaoka, Design of intelligent artifacts:a fundamental aspects, Proc. JSME International Symposium on Optimization and Innovative Design(OPID97), pp , September 1997 [] Olli Seppanen, William J. Fisk: A Model to Estimate the Cost- Effectiveness of Improving Office Work through Indoor Environmental Control, Proceedings of ASHRAE, [] Peter R. Boyce, Neil H. Eklund, S. Noel Simpson, Individual Lighting Control: Task Performance, Mood, and Illuminance, Journal of the Illuminating Engineering Society, pp.11-1, Winter [] M.Miki,T.Hiroyasu,K.Imazato,Proposal for an intelligent lighting system,and verification of control method effectiveness, Proc IEEE CIS, pp.-, [6] M. Miki, T. Hiroyasu, and K.Imazato, Proposal for an intelligent lighting system, and verification of control method effectiveness, Proc. IEEE CIS, pp.-, [7] S.Tanaka,M.Miki,T.Hiroyasu,M.Yoshikata,An Evolutional Optimization Algorithm to Provide Individual Illuminance in Workplaces, Proc IEEE Int Conf Syst Man Cybern, Vol., pp.91-97, 9