Applicability of Advanced Light Control Concepts with KNX

Transcription

1 KNX Scientific Conference, Las Palmas, Gran Canaria/Spain, Applicability of Advanced Light Control Concepts with KNX Manfred Mevenkamp Institut für Informatik und Automation, Hochschule Bremen Flughafenallee 10, Bremen Introduction and main results The energy savings potential of the lighting sector in Europe is estimated to be about 40%. For commercial buildings even higher efficiency improvements up to 55% can be achieved [1]. These energy savings can mainly be realised by using energy efficient equipment, e. g. efficient luminaries with electronic ballasts, and by intelligent light control systems which evaluate human presence and minimize artificial light in the daytime. Solutions and products for daylight-dependent light control are commercially available. The present study shows that not all of these provide adequate control results in all operating conditions and that there is still quite a potential for improvements with respect to control performance and cost effectiveness. Control strategies using only one sensor to control three groups of lamps aligned parallel to a window front have been analysed experimentally and by simulation. It is shown that master-slave strategies fail to work properly under certain lighting conditions. Proper light control requires additional information about light distribution and luminaire characteristics. Figure 1: Daylight distribution and artificial light controlled by a single sensor The new model-based control strategy presented in [2] is shown to provide good control performance under all lighting conditions, even in an implementation with certain simplifications. Applicability of the concepts covered in this study is discussed with respect to performance and commissioning complexity. Modifications necessary for masterslave concepts to handle certain situations are presented and for the model-based control system a practically viable commissioning procedure is described.

2 1 Single sensor daylight dependent light control The artificial light necessary to comply with given lighting requirements (e.g. [3]) depends on the distance from the windows and on the daylight. In commercial buildings we frequently find large rooms with three or more groups of lamps aligned parallel to a window front. Light control concepts for rooms with only one or two luminaires or luminaire groups cannot easily be carried over to this larger type of room. Possible daylight dependent light control strategies for such a room can roughly be classified as follows: 1. Light control with outside illuminance sensor 2. Light control with dedicated sensors for each group of lamps 3. Master-Slave concepts using only one sensor Figure 2: Daylight dependent light control can be especially efficient in seminar rooms. 4. Model-based control using one sensor and illuminance estimation The first one is not feasible in rooms with sun blinds, which are frequently used in commercial buildings. The second one, though quite common today in small office rooms, will generally be too expensive in larger rooms. Therefore, the single sensor concepts with master-slave strategies or model-based designs are the most promising solutions in this case. 1.1 "Offset" Master-Slave Control Master-Slave control strategies use the sensor measurement value to control one group of lamps, the "master", which is typically the one in the centre of the room. Given the dim level of this master group, offset-type master-slave control systems set the dim levels at the windows and at the opposite wall to values which differ from the master level by certain fixed offsets that are configured during commissioning. This rather "ad hoc" approach is at present the only commercially available single sensor control system for more than two groups of lamps. Figure 3 visualizes the dependency between slave- and master-levels. Two critical situations can be identified, where the slave levels cannot be set to suitable values based on the master dim Figure 3: Typical slave dim level setting by Offset-type Master-Slave concepts - 2 -

3 level alone. As the master level approaches 100% (with decreasing daylight) there is no information on how and when the dim level of the window group should also be increased towards 100%. And when the master group is turned off (with increasing daylight) the concept provides no information to decrease the dim level at the wall in a well defined way. These deficiencies were demonstrated in a testimplementation using a commercial KNX sensor with an offset-type masterslave light control application (figure 4) [4]. Experiments with increasing and decreasing daylight Figure 4: Sensor and luminaire positions in laboratory studies with different choice of offset values resulted in illuminance levels which did not comply with the given requirements. Figure 5 shows illuminance values measured on the laboratory desks on a rainy day at noon and towards the evening. The dim level of the master group is controlled to match the reference value (500 lx). The lamps at the wall are in both situations fully turned on even with the lowest chosen offset of +20%. Near the windows the required illuminance at noon is maintained even with an offset of -60%. However, in the evening the dim levels of the lamps near the window remain too low. With an offset of -60% only 200 lx were measured. Figure 5: Illuminance with Master-Slave-Control at noon (left) and in the evening (right) On the other hand, when there is sufficient daylight at the location of the sensor, the the master group is switched off. As a consequence the sensor used in this test switches off all the slaves, too. This results in insufficient lighting at the wall opposite the windows and with changing daylight also repeated switching of lamps can occur. From this it can be stated, that offset-type master slave concepts in this simple form are not properly applicable for constant light control with three or more groups of lamps

4 1.2 "Proportional" Master-Slave Control Another Master-Slave control strategy that will be commercially available in the near future uses certain factors instead of offsets to set the slave dim levels. It starts with initial dim levels suitable for the daylight of a cloudy day (e. g. 50%, 80% and 100% in figure 6). With increasing or decreasing daylight the increments or decrements of the slave dim levels are chosen proportional to the change of the master level. The dependency between slave- and masterlevels visualized in figure 6 shows that this concept handles master levels approaching 100% in a well defined way. But with increasing daylight there still remains a critical situation. As soon as the master group is turned off, this concept, too, provides no information how to suitably decrease the dim level at the wall. The proportional master-slave scheme has been investigated by simulations with a well proven light distribution model of the laboratory. As shown in figure 7 the "Proportional" design performs better than "Offset" control (here with offset = 30%) but the result is still not satisfactory. Illuminance near the windows does not reach the required level of 500 lx. Figure 6: Typical slave dim level setting by a "Proportional" Master-Slave concept The main reason in this case is a nonlinear relation between dim level and illuminance measured on the desks. Master-slave concepts and the offset design in particular rely on the assumption of a linear relation between dim level and illuminance, which may be common sense but is not necessarily true in general. Figure 7: Performance of Master-Slave-Concepts in the evening The luminaire dim characteristics measured in the laboratory for each of the luminaire groups are shown in figure 8. They are almost identical and can be approximated with only small deviations by a third order polynomial. With a dim level of 80% these luminaires produce only 55% of their maximum illuminance. Control designs set up with a linear relation between dim level and illuminance in mind cannot be expected to work properly with such a type of luminaire

5 As a conclusion master-slave designs are not applicable in their simple, "ad hoc" form. The algorithms have to be adjusted using additional information about 1) dim level characteristics of the luminaires to suitably set the slave parameters and 2) daylight distribution in the room to cope with the situation when the master dim level becomes zero. Figure 8: Luminaire dim characteristic This means that in any applicable single sensor light control system for three or more groups of lamps some kind of model of the lighting characteristics of the room must be incorporated. 2 Model based daylight dependent light control 2.1 Control concept and estimation algorithm With a full model of light distribution and luminaire characteristics in a room the illuminance below each group of lamps can be estimated and compared to the reference value to yield a control deviation. This way all groups can be controlled separately using the same algorithm as the master group in a master-slave system. Such a concept was first presented in [2]. The estimation is based on calculating the illuminance, of each group from its dim level via the characteristic, = (see above, figure 8). The daylight at the sensor position, then equals the difference between the sensor measurement value and the illuminance contribution from the luminaires. If we describe the contribution of group to the illuminance at the position of group by a distribution factor,, then the total illuminance near the groups is where: t = L + 0 0, i, M L = L ( ) E D E D E E F L = Vector of total illuminance values at the group positions, = Matrix of distribution factors, (, = 1), = Vector of illuminance values of all groups (on the desks below), = Index of the group where the sensor is located,, = Daylight near the sensor position, = Vector of daylight distribution factors (, = 1), = Vector of dim levels of all groups, [ 0, 1] L. i - 5 -

6 , is the sensor measurement value which means that, can be calculated from the -th row of the above vector equation. In the special case of three groups of lamps with the sensor near the centre group ( = 2) the implementation of the algorithm is quite simple. For the system tested in the laboratory only light contributions from neighbour groups have to be taken into account (i. e., =, = 0). So the full distribution model consists of only 6 parameters that have to be determined by measurements during commissioning. Once the total illuminance values are known any common constant light control algorithm can be used to set the dim levels. 2.2 Implementation and Test In the present system the controller changes the dim level of each group according to the deviation of its estimated illuminance from the setpoint (e. g. 500 lx) via a ramp function (integrator with limited input). The ramp steepness defines the control speed and has to be chosen carefully taking ergonomic aspects into account. Especially small daylight changes should not lead to noticeable changes of artificial light. On the other hand a faster reaction is needed when large sudden changes occur, e. g. when the blinds are shut. This is realised by a variable steepness of the ramp function depending on the actual setpoint deviation. The complete algorithm was implemented on a PLC within the KNX system of the laboratory. Distribution matrices = and = for artificial light and daylight were used in the estimation model. Because of the inhomogeneous light distribution in the room the reference value was set to 550 lx to ensure sufficient lighting at all positions. The dynamic and stationary performance was evaluated by gradually shutting and opening the blinds at daytime. The result is shown in figure 9. The upper diagram shows the sensor measurement value and the estimated value of daylight at the sensor position. As can be seen from the lower diagrams daylight is sufficient near the windows in this test. Illuminance measured on the desks between the windows and the centre is near 600 lx at first, the estimated illuminance at the windows is about 900 lx. Therefore the window group initially is dimmed to zero (lower diagram). Blinds operation starts after 30 seconds. The control system increases the centre group dim level as soon as the sensor measurement falls below the reference value. At t = 45s the estimated illuminance at the windows reaches the reference level and rapidly goes far below it. Therefore the controller increases the dim level of the luminaires at the windows using a steeper ramp than in the centre. After about 90 seconds the blinds are shut completely and dim levels remain almost constant until after 130 seconds first the slats are turned and then at t = 160s the blinds are opened again. Dim levels and illuminance then return to the values of the initial situation

7 Figure 9: Control results of the model based concept with closing and opening blinds The change of light caused by blind operation is so fast that illuminance in the room temporarily goes down to 300 lx. The controller dynamics might be adjusted to allow for a faster reaction. However, as mentioned above this should be done carefully so that light changes do not become perceivable or even disturbing. In stationary situations and with slow daylight changes the control system performs very well. 3 Applicability commissioning and performance Easy and foolproof commissioning is an important aspect for successful light control applications. At least the commissioning process should be straightforward and well defined. For any light control system this process involves parameter settings and measurements, at least illuminance measurements for sensor calibration. 3.1 Master-slave designs Simple Master-Slave devices only need a few parameters to be set during commissioning (e. g. offset values) but clear guidelines for suitable settings are not always available. Often a more or less heuristic choice is made so that an acceptable performance is not guaranteed

8 Moreover, at this time the abovementioned problems of master-slave concepts are not yet solved. Enhancements to the algorithms are needed which evaluate the sensor measurement value when the master dim level is zero. By this the dim level at the wall can also be reduced towards 0% if the measured value indicates abundant daylight. Obviously some knowledge about daylight distribution would have to be incorporated in this calculation. In the case of "Offset" concepts also the situation when the master level is 100% has to be considered. Possible solutions to systematically increase the dim level at the windows in this situation are not straightforward and involve special positioning of the sensor or knowledge about the contribution of this group of lamps to the illuminance at the sensor position. Alternatively this case would have to be handled by manual operation which, of course, would not be a preferable solution. There seems to be also no straightforward way to integrate special dim characteristics of luminaires into master-slave designs. Their performance therefore depends on the linearity of this characteristic which may be critical especially with offset-type designs. Summing up, commissioning of master-slave light control systems becomes more complex if the concepts are refined to handle all light situations properly. Commissioning then implies additional parametrization and extra information to be obtained by measurement. 3.2 Model based light control Parametrization of the model based light control system involves illuminance measurements to determine the luminaire dim characteristic and the distribution factors, of the estimation algorithm. Sensor calibration is an integrated part of this process. The procedure consists of 3 steps: 1.) Measurement of illuminance below the centre group with other groups switched off and no daylight (blind shut) at dim levels L = 20, 40, 60, 80, 100%. With L=100% also measure illuminance and below the neighbour groups. Let = and = % and choose!"#!$!"#!$,& = 0.9, &, = 1.1,,& = 0.9, &, = 1.1. (The factors 0.9 and 1.1 ensure sufficient artificial light even in case of model deviations.) 2.) Sensor calibration following the manufacturer's instructions. 3.) Measurement of illuminance at daytime (blinds up, all lamps switched off) below each of the luminaire groups. Let ),, ),& and ), be the measured values, then, = *, *,+ and, = *,% *,+. These are about 10 measurements, which is more than master-slave designs require. But the procedure is straightforward and results in a system that eliminates the problems that are inherent in master slave concepts

9 4 Conclusion The model based daylight dependent light control method presented here is well applicable with respect to performance and commissioning complexity. Commissioning may be facilitated be pre-defining distribution factors for certain types of rooms or rooms with given dimensions. This will be part of further research. Also, measurement deviations of sensors and their influence on light control performance have to be investigated in more detail. Daylight dependent light control plays an important role in all energy efficiency scenarios (see e. g. [5]: "Energy Efficiency with KNX"). The limited performance of existing solutions is hardly supportive for these goals. Work on refined systems with good acceptance by the users is therefore advisable. References [1] Schornick, D.: Energy saving potentials in lighting, REHVA Journal, 01/2011 [2] Mevenkamp, M.; Cholbi Collados, M.: Model-based daylight-dependent light control using KNX, KNX Scientific Conference, Pamplona, 11/2010 [3] DIN EN 12464, Light and lighting Lighting of work places, Part 1 "Indoor work places", 2003 [4] Cholbi Collados, M.: "Cost-effective daylight-dependent light control of several groups of lamps", Bachelor-Thesis, Bremen University of Applied Sciences, Bremen, 2010 [5] %20Flyer/English/KNXEEnew_EN_Screen.pdf - 9 -