PREDICTIVE MODEL-BASED CONTROL OF VENTILATION, LIGHTING, AND SHADING SYSTEMS IN AN OFFICE BUILDING. Vienna University of Technology, Austria

Transcription

1 PREDICTIVE MODEL-BASED CONTROL OF VENTILATION, LIGHTING, AND SHADING SYSTEMS IN AN OFFICE BUILDING Matthias Schuss 1, Claus Pröglhöf 1, Kristina Orehounig 1, Sokol Dervishi 1, Mario Müller 2, Heinz Wascher 2, an Areshir Mahavi 1 1 Department of Builing Physics an Builing Ecology,, Austria 2 Hans Höllwart - Forschungszentrum für integrales Bauwesen AG, Stallhofen, Austria ABSTRACT This paper reports on ongoing work towar implementing a preictive control approach for builings systems for ventilation, lighting, an shaing. The main objective of this metho is the optimize control of multiple evices towar usage of passive cooling an natural lighting. Thereby, control options (various opening positions of winows, shaes, etc.) are generate an computationally assesse using a combination of option space navigation via genetic algorithms an numeric simulation. INTRODUCTION In the last few years system an energy expenitures for space cooling have ramatically increase, even in central-european climatic zones. This has encourage the efforts to evelop an implement smart (energy-efficient) cooling methos. An intelligent control approach involving all relevant systems an enowe with the capacity of proactive (preictive) control is believe to have the potential to signicantly reuce builings' energy eman. Towar this en, passive cooling, avance shaing control, an increase usage of natural light is essential. Possibilities to use natural ventilation an builing controls in existing builings were presente in previous publications (Mahavi & Pröglhöf 2004, 2005, an 2006; Mahavi et al. 2008; Orehounig 2010; Pröglhöf 2010). This paper further evelops a new simulation-base preictive control approach (Mahavi 2008; Mahavi et al. 2009) with the capability to facilitate the application of the aforementione sustainable inoor climate control systems. The core iea behin this approach is the use of numeric builing performance simulation applications to preict ahea of an actual control action the consequences of multiple control options. Once the options are generate an virtually realize via simulation, they can be evaluate an ranke, thus proviing a basis for optimal control ecision making. METHOD To implement the propose moel-base control strategy a realistic setting is essential. Therefore, we selecte two builings for implementation. This paper focuses on one of these builings, namely a moern office builing ("Fibag") in Stallhofen, Styria, Austria (see Figure 1 to Figure 3). The builing has a typical glass an aluminium façae (Figure 1). The primary structure is massive (concrete skeleton, floors, an stcases), but the internal (partition) walls may be escribe as lightweight. Two test rooms in this builing were selecte for experiments. One room was use to test the control approach (see Figure 1 an Figure 3), whereas the secon room was use as a reference room. The two test rooms are ientical layout-wise an are locate in the first an secon floor above each other, facing north an east irections. The builing is locate in a rural, low-ensity, an low-rise context. Figure 1. The Fibag Builing Builing Performance Simulation in a Changing Environment - A. Mahavi / B. Martens (es.) - 377

2 Figure 2. The test room To emonstrate the feasibility of the simulation-base control approach in a multi-system context, sensors an actuators were eploye: The rooms are equippe with programmable room controllers, inoor environmental sensors (Figure 4), as well as actuators for the automate operation of winows (Figure 5) an blins. Moreover, to monitor local climatic conitions, a weather station (Figure 6) was installe on top of the builing. Table 1 provies a escription of all system components. A schema of the test system is illustrate in Figure 4. Figure 3. Floor plan of the test room Table 1 System components escription SYSTEM COMPONENT Inoor climate sensors Outoor climate sensors User action an presence sensor Shaing evice Controllable lighting system Backbone an communication network DESCRIPTION Compact inoor climate stations measuring temperature, relative humiity an velocity as well as carbon ioxie an raiance Weather station measuring temperature, relative humiity, precipitation, global irraiance, win spee an win irection. Presence: PIR - Sensor with settable threshol time Door opening: magnetic contact sensors Two synchronize sleepless settable rives for each winow to control the winow opening position continually Single rives with a special gear unit for height an angle positioning The room controller coul set on/off an imming levels between % of the total lighting power IP base communication with access to builing ata points an ata history Figure 4. Internal sensor station Figure 5. Automate Builing Performance Simulation in a Changing Environment - A. Mahavi / B. Martens (es.) - 378

3 ZONE 1 ZONE 2 RH CO2 E WC OC vi g S RH CO2 E WC OC vi g S Controller T2 (RESI) Controller T2 (RESI) Blin 4 Blin Blin 2 3 Blin Blin 4 Blin Blin 2 3 Blin TCP/IP TCP/IP LAN WS Server RESI Visualisierung Trens DB Server Controller application DB Simulation Matlab Figure 7. Schema of the test system Figure 6. Weather station The aforementione moel-base control approach is being implemente in the test room. Thereby, weather forecast information(weather.com, 2010) is fe into simulation applications to regularly probe the implications of various control action alternatives in view of esirable inoor-environmental conitions. Thus, the likely optimal course of control action can be ientie proactively towar optimization of energy an environmental performance of the builing. An essential avantage of the propose approach is its ability to consier the thermal storage capacity of the builing's thermal mass more reliably. In orer to better ocument the performance of the implemente control regime, we will use the secon room as a reference room for comparison. Preictive Control approach The present paper attempts to further evelop the preictive control approach (see Figure 8 an Mahavi 2008). Instea of the previously applie combination of the greey search metho (combine with stochastic jumps), we know explore the potential of genetic algorithms towar navigation of the control options search space. This moication is necessary, since we woul like to be able to generate an evaluate control options on a regular basis (i.e. in short time intervals). Moreover, temporal changes in the position of evices over time (operation scheules) must be consiere for each simulation run. These leas to an explosion of the control options (schemes), which coul be better tackle via genetic algorithms. Thereby, weather forecasts (Weather.com 2010) together with expecte values require for simulation input (e.g. internal gains) are the starting point for a series of multi-omain simulations (thermal an lighting) base on a genetically prouce variation of alternative control states. The control process was implemente in MATLAB (MATLAB 2010) environment. The implementation eploys HAMBase (van Schijnel 2007) an Raiance (Raiance 2010) as incorporate simulation tools. Builing Performance Simulation in a Changing Environment - A. Mahavi / B. Martens (es.) - 379

4 Bounary conitions at ti Control systems state at ti Room 1 Device 1 Device 2 Name State Device n Physicalaress Forecasting moule Alternative states Room 2 Device 1 Device 2 Device n Preicte bounary conitions for the interval [ti, ti+n] Multi-omain simulation engine Room n Device 1 Device 2 Preformance Inicators Preicte performance for the interval [ti, ti+n] Device n Figure 9. Schema for evice attribute efinition ata structure Control systems state at ti+1 time Optimum control systems state at ti+1 Base on the first generation simulation the bestranke scheules were selecte to generate new chil scheules in a ranom multipoint crossover reprouction process (Figure 10). For this purpose, the high-ranke scheules are crosse with themselves as well as with aitional ranomly selecte scheules ealing as parent elements. Figure 8. Illustration of the preictive simulation assiste control strategy Ranom bit Pattern Parent scheule 1 Parent scheule 2 chil scheule These simulations results are the basis for the evaluation process to generate optimum control ecisions accoring to efine performance inicators. This preictive control approach operates in ference to the commonly use reactive feeback base stanar control methos use in builing systems control. Instea of using ferences of the set values an actual values, this approach optimizes the system operation in a holistic way. Alternative States To fee the preictive control metho with alternative operation states, the relevant evice control scheules have to be prouce. The generative process of scheules uses genetic algorithms. A number of efault operation scheules are use together with ranomize scheules as the initial setup. Neee state efinitions an evice attributes are store in a preefine ata structure (Figure 9) to generate the scheule automatically. + = Figure 10. Illustration of the genetic multipoint crossover reprouction The ranking is one by a number of performance inicators (iscusse below) to estimate the fitness of each alternative state. Figure 11 illustrates this genetic approach. Builing Performance Simulation in a Changing Environment - A. Mahavi / B. Martens (es.) - 380

5 Iniviuals: Generation: 1 t i n ( k ) k t i (2) n Multi omain Simulation Best fitness ranking Creation of net generation m=2 The calculation of each eviation epens on a fixe set point or an acceptable parameter range as shown in Figure 12. The general inicator i x coul be erive either linearly (Figure 13), or exponentially (Figure 14). The principle calculation proceure for power respectively energy inicators is presente in Figure 15 an expresse for HVAC an lighting power use. Iniviuals: Generation: m p n p SP Multi omain Simulation Best fitness ranking Creation of net generation m=m+1 ti ( t) p( t) p SP (t ) p t ti+n Best performing scheule p max Figure 11. Illustration of the genetic generation of the esire operation scheules Performance Inicators A holistic evaluation of alternative system operation scenarios with relate control system states is the core component of this control metho. A set of builing performance inicators weighte with associate weighting factors were use to evaluate the multi-omain simulation results an rank the alternative control state scenarios. The performance inicator i (Equation 1) is the weighte sum of all inicators i x. The value of each inicator an the sum of the weighting factors w x is in the range of 0 to 1. Hence i must be in the same range. The ranking of the alternatives is one by maximum to minimum sorting. p min ti p min p ( t ) p ( t ) p min ( t ) 0 p min p ( t ) p max p ( t ) p max p ( t ) p max ti+n Figure 12. Deviation calculation for a general system parameter p. i 1 t i i x w x (1) 0 max i,i x, w x 0,1 an w x 1 The calculation of each inicator is base on the simulate preictive tren of the relate system parameter (e.g. temperature of a room). For each specic parameter, the sum of eviations is calculate for the future n time steps shown in Equation (2). x i x 1 0 max max max Figure 13. General linear performance inicator i x calculation. Builing Performance Simulation in a Changing Environment - A. Mahavi / B. Martens (es.) - 381

6 i x i e c. simulations in EDSL Tas (EDSL 2008). Figure 17 shows the external temperature e an the simulate inoor temperatures i for a typical summer week. The simulation was one for an change of 0.4, 1.4 an 10 h -1 over 24 hours a ay. A ventilation regime with an change rate of 0.4 h -1 over the ay (8am to 7pm) an 10 h -1 uring the night hours was simulate as well. These simulations inicate the overheating tenency of the rooms, but also showe the potential of natural ventilation. Figure 14. General exponential performance inicator i x calculation. i PHVAC 1 0 i PHVAC 1 n t i n t t i P HVACmax P HVAC 1 P HVAC ( t ) P HVAC max Figure 16. Measurements in first floor test room for July t i n i PL 1 P Lighting ( t ) 1 n P t t i Lighting max Figure 15. Performance inicator for power or energy relate parameters, as expresse for HVAC or Lighting relate power use P HVAC an P Lighting. RESULTS Data is being collecte in both test rooms towar an objective ocumentation of the inoor-environmental conitions. To obtain an initial impression regaring the impact of winow ventilation on inoor temperature, measurements of the external temperature e [ C], the test room s temperature i [ C], an the winow opening pos w [%] are shown in Figure 16 for a typical summer week. Thereby, the influence of two instances of (manual) winow operation can be seen. Both rooms have a very strong overheating tenency cause by the limite thermal mass an the oversize winows. The usual summer ay temperature is in the range from 20 to 30 C with peaks up to 35 C. Simulate natural ventilation Parallel to the monitoring phase, thermal simulations were one to estimate the night cooling effect an virtually test the new control approach. For this purpose, measurements of change rates were the starting point for ferent natural ventilation Figure 17. Simulate inoor temperatures as a function of change rates Control approach Implementation To emonstrate the avantages of the preictive control, a first implementation was one in a virtual setup. Base on the HAMBase simulation package (van Schijnel 2007) for MATLAB a thermal moel of the two test rooms was create. Aaptations to HAMbase were carrie out for the control of shaing an the possibility to run single hour step simulations with store ata. The evelopment an integration of the complete control system was also one in MATLAB. Components for the collection of require ata (weather forecast, internal/external sensor ata) an their storage into a sqlite atabase were programme in C. These coul be run inepenent of the control program as a service. Builing Performance Simulation in a Changing Environment - A. Mahavi / B. Martens (es.) - 382

7 At this stage only the room temperature was use as a performance inicator. The comfort zone for the room temperature was assume to be the range between 20 an 25 C (Figure 18). -max =25 -min =20 t t i t i +n ( t ) 0 min ( t ) - max ( t ) ( t ) - min ( t ) min max ( t ) - max Figure 20. Temperature of the test room (θ i,sim ), :00 Figure 18. Deviation calculation for the room temperature Result of a test using measure external climatic ata an the HAMbase moel is presente in Figure 19, Figure 20, an 22. Each plot shows the historical ata incluing the real external temperatures an the simulate inoor temperatures on the left half of the plots. Simulate temperatures for all scenarios (generate via the aforementione genetic approach) are presente in grey color on the right half sie together with the status of winows (green) an shaes (blue) for the best performing scenario (black). Concerning the status scale, 1 enotes fully open winows an fully close shaes. These Figures represent 3 consequent ays. They illustrate the large ference between weather forecast an actually measure temperatures. However, the performance of the system (i.e. ientication of the best performing scenario) oes not appear to be aversely affecte by such weather forecast errors. Figure 19. Temperature of the test room (θ i,sim ), :00 Figure 21. Temperature of the test room (θ i,sim ), :00 DISCUSSION The scope an the initial results of a prototypical implementation of a simulation-assiste preictive control approach for passive cooling were presente in a recently constructe office builing in Austria. Thereby, the potential of the metho was primarily explore towar harnessing natural ventilation (via winow elements equippe with software-controlle actuators) an solar control (via automate shaing evices). The results thus far point to the potential of the propose control metho, which involves the ynamic an parametric use of numeric simulation of genetically generate alternative control options to proactively assess, compare, an evaluate control these options towar ientication of the control actions that yiel appropriate inoor-environmental conitions while minimizing energy use. Future efforts will focus on the long-term test an monitoring phases in occupie settings. Builing Performance Simulation in a Changing Environment - A. Mahavi / B. Martens (es.) - 383

8 ACKNOWLEDGEMENT The research presente in this paper is supporte in part by funs from FFG "Naturally Cool" (Project- Nr: ) as well as the K-project "Multunctional Plug & Play Façae". REFERENCES EDSL, A-TAS Version 8.5, Environmental Design Solutions Limite, Mahavi, A., Pröglhöf, C Natural ventilation in builings Towar an integrate control approach, Proceeings of the 35th Congress on Heating, Refrigerating an Air-Conitioning, Belgrae, Serbia, pp Mahavi, A., Pröglhöf, C A moel-base metho for the integration of natural ventilation in inoor climate systems operation, Proceeings of the 9th International IBPSA Conference, Builing Simulation 2005, Montreal, Canaa, pp Mahavi, A., Pröglhöf, C A moel-base approach to natural ventilation, Builing an Environment, Elsevier, Volume 43(4), pp Mahavi, A Preictive simulation-base lighting an shaing systems control in builings, Builing Simulation, an International Journal, Springer, Volume 1, Number 1, ISSN , pp Mahavi, A., Orehounig, K., Pröglhöf, C A simulation-supporte control scheme for natural ventilation in builings, Proceeings of the 11th IBPSA Conference, Builing Simulation 2009, Glasgow, Scotlan, pp MATLAB, MATLAB Release 2010a, The MathWorks, Inc., Orehounig, K., Mahavi, A., Pröglhöf, C., Schuss, M Virtual implementation of a simulationassiste passive cooling strategy in builings, Proceeings of the 10th Rehva Worl Congress, Sahin N. (e.), Antalya, Turkey. Pröglhöf, C., Schuss, M., Orehounig, K., Mahavi, A Incorporation of a novel passive cooling metho in an existing builing, Proceeings of the 10th Rehva Worl Congress, Sahin N. (e.), Antalya, Turkey. Raiance, Raiance Synthetic imaging system Version 4, University of Calornia, Schuss, M., Pröglhöf, C., Orehounig, K., Mahavi, A A case stuy of moel-base ventilation an shaing controls in builings, Proceeings of the 10th Rehva Worl Congress, Sahin N. (e.), Antalya, Turkey. van Schijnel, A.W.M Integrate heat an moisture moeling an simulation, PhD thesis, Einhoven University of Technology, or accesse June Weather.com, National an Local Weather Forecast, Hurricane, Raar an Report, The Weather Channel Interactive, Inc., Builing Performance Simulation in a Changing Environment - A. Mahavi / B. Martens (es.) - 384