Advertisement

48 views

SENSOR-BASED INFORMATION MODELING FOR LIFE CYCLE COMMISSIONING OF RESIDENTIAL BUILDINGS

SENSOR-BASED INFORMATION MODELING FOR LIFE CYCLE COMMISSIONING OF RESIDENTIAL BUILDINGS
of 8
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Share
Tags
Transcript
  Proceedings: Building Simulation 2007 - 1572 - SENSOR-BASED INFORMATION MODELING FOR LIFE CYCLE COMMISSIONING OF RESIDENTIAL BUILDINGS Yun-Shang Chiou and Khee Poh Lam Department of Architecture, Carnegie Mellon University Pittsburgh, PA 15213, USA ychiou@andrew.cmu.edu ABSTRACT According to the National Institute of Building Science (2006), one of the main goals of building commissioning is to “maintain facility performance across its life cycle”. In recent years, the maturing of sensing technology has helped to advance this vision through sensor-assisted building commissioning. For residential projects, however, the overhead associated with long-term sensor performance and quality assurance and sensor data analysis creates a barrier that diminishes the benefit of such technological advancement. This paper describes how the principles of complex system theory can be applied to address these issues in the context of residential building. Two cases are  presented. First case demonstrates the application of “scale” and “format” that underline the idea of “bounded rationality” in dealing with imperfect sensor data. Second case refers to the idea of “decomposable hierarchical system” as the framework for building system information modeling to reduce the level of difficulty for resident’s  participation in life cycle commissioning. KEYWORDS Information Model; Life Cycle Commissioning; Residential Building; Sensor-Assisted; Complex System Theory INTRODUCTION In recent years, the maturing of sensing technology has helped to advance the role of building commissioning in “maintain facility performance across its life cycle” through sensor-assisted continuous commissioning. Several studies of sensor-assisted building commissioning have been conducted under non-residential context. Piette (2000) evaluated a prototype Web-based Information Monitoring and Diagnostic System (IMDS) that covers main components of the HVAC system in an office building through 57 high-quality sensors. The IMDS is used for control problem and equipment fault detection. Piette estimated the yearly saving in  building operation is worth $20,000 - enough to pay off the IMDS in about 5 years. Wang (2002) developed a strategy based on generic algorithm to automatically diagnose and evaluate the sensors of  building refrigeration systems during initial and continuous commissioning to assure the accuracy and reliability that are essential to the effectiveness of the  building management system (BMS). Singhvi (2005) deployed a pervasive temperature sensor network to determine the effects of room usage on building conditions. The results show that the building state is dynamic and the indoor environmental parameters vary significantly based on occupancy and layout of the room. The results also suggest that for collecting  building performance data, it requires a pervasive, rather than sparse, sensor network. These case studies suggest that cost of system monitoring and diagnostics, reliability of sensor data and the metrics of sensors to capture sufficient information are three key factors to the success of sensor-assisted building commissioning. A series of studies (Matson 2002, Wray 2002, Wray 2000) have identified the metrics and benefits of residential building commissioning. However, the issue of commissioning cost was not specifically addressed. Earlier, Westergren (1999) conducted research with continuous sampling of whole house energy consumption data and applied statistical method to construct a single-family energy consumption model. Yet the reliability of the monitoring data has not been discussed at all. Literature review indicates that there is a lack of research focus on issues related to sensor-assisted commissioning in the context of residential building. Figure 1. InfoMonitors web-interface database (www.infomonitors.com/IBACOS)  Proceedings: Building Simulation 2007 - 1573 - Two sets of detailed annual sensor data are acquired from IBACOS InfoMonitors database (Fig. 1) for this study. The first data set is from an occupied residential house in Aspen, Colorado equipped with PV (Photovoltaic) system. The second data set is from an occupied residential house in Buffalo, New York, equipped with CHP/FC (Combined Heat and Power /Fuel Cell) system. Sensor data include  parameters such as room temperature, humidity, solar radiation, total and circuit by circuit electricity consumption, water flow and temperature, and natural gas consumption (Fig. 2, Fig. 3). Figure 2. Front page of Aspen house in InfoMonitors Figure 3. Sensor data graphs generated by InfoMonitors Sensor data quality remains the primary challenge and concern. System setting records and on-site investigation indicate that quality issues embedded in the sensor data include system performance, occupant life style, sensor accuracy and precision, data capturing schedule settings, and data organization. Due to the heterogeneous nature of these causes, noise in data stream is difficult to filter out through conventional top-down reductionism methodology. “Bounded rationality” from complex system theory (Simon 1996) is adapted in this study to deal with the  problem of data analysis. This concept provides an alternative view to look at the noise in sensor data and guides the definition of the boundaries of multi-level views for the design of the sensor data embedded hierarchical information model. “Bounded rationality” can be interpreted as “human rationality is bounded by the accessibility of relevant information”. The value of information depends largely upon its accessibility to the decision maker. The accessibility of information has much to do with the format of presentation. The assertion of this study is, even when the quality of data is less than perfect or the sensor data is ambiguous in meaning, if it is  presented in a form appropriate for the designated decision context, information derived from sensor data can still lead to high-quality decision making. Different issues occur during building commissioning which can be result from different  problem domains as well as problem resolutions. By adopting principles of “complex system theory”, characterized by decomposable hierarchical structure (multi-scale model structure) and bounded rationality (expertise-oriented product unit), sensor data are  processed and transferred into “accessible” information format. Through this approach, this study shows that the seemingly interwoven yet heterogeneous issues can be disentangled and categorized into knowledge domains in respective hierarchies, thus greatly reducing cost and knowledge overhead that prevents the implementation of continuous commissioning in residential scale. In the Aspen house, the objective of data analysis is to identify opportunities in reducing on-site electricity consumption. Data analysis shows that, regardless of the uncertainty of the accuracy and  precision of sensor data, the resulting graphs, when formatted correctly, still provide highly useful information on how to achieve this objective. Sensor data from Buffalo house is used to construct the information model to monitor the house’s operational energy efficiency. The information model demonstrates the use of “multi-level information modeling” approach (transforming sensor data to hierarchical information structure of system, sub-system and components) for building operation fault detection, system diagnostics and knowledge discovery. It has greatly simplified the knowledge requirement for system diagnostics thus enabling the occupants to assume the daily role of the facility manager for residential building. SENSOR DATA PROCESSING Aspen house electricity consumption analysis  Proceedings: Building Simulation 2007 - 1574 - IBACOS InfoMonitors database collects Aspen house sensor data from December 2004 through April 2006. The Aspen house project contains 19 room and equipment temperature (F) channels, 3 relative humidity (%) channels, 35 circuit-by-circuit (Amps) channels, 2 equipment runtime channels, 4 water flow (Gal) channels, 3 whole system electricity (PV gain, to Grid, from Grid) (kWh) channels, 4 heat output (BTU) channels, all recorded in 15 minute time step. The electricity consumption sensor data is extracted from the circuit-by-circuit channels and the whole system electricity channels. Data are processed in two ways. First, Amp readings of all 35 circuits are converted to kWh unit and summed up to compare against the readings of the sum of whole system electricity 3 channels to verify the consistency of the sensor data. Second, 35 circuits are ranked by their electricity consumption. To understand the patterns of electricity consumption of the key circuits, data of top 10 circuits are  presented in graphs for further analysis. Two pieces of external information namely occupancy and event changes during the period of data collection are used to assist the circuit-by-circuit sensor data processing. A working couple and their young baby live in the Aspen house. In expectation of the significant difference of occupancy patterns between workday and non-workday, each “top 10” circuit has a graph with pair of curves derived from weekday/ weekend data. Second, sensor data provider (IBACOS) has verified several sensor metrics anomalies and changes of occupancy pattern during the one year experiment period. Additional graphs are generated for the observation of the effects of these documented events. Buffalo house energy conversion efficiency information model The Buffalo house energy conversion efficiency information model is implemented in two steps: information schema design and the proof-of-concept web-based information model deployment. Its goal is to enable occupants to monitor the performance and to pinpoint the problem of the system with minimum training. On the information schema design, all sensor points on the heat and electricity system diagram are associated with corresponding sensor channels in the data set. From the system diagram, a new IDEF0 hierarchical system diagram is constructed based on  process unit and its resolutions. With the guidance of the new IDEF0 diagram, the sensor data are organized to provide “views” of the state of the respective process unit and energy flow. On web-base information model deployment, a mirror database of all sensor data is created using a local MySQL database management system. General Java programming tools linked to the MySQL database enable data organization and processing. Finally, the IDEF0 information schema is placed on a tomcat web server with Java Servlet connected to MySQL with Java 2D graph library to implement the Web-based information model. Electicity Load Error between in-house circuits and whole house meters -30.00%-25.00%-20.00%-15.00%-10.00%-5.00%0.00%5.00%10.00%15.00%20.00%Nov-04 Feb-05 May-05 Sep-05 Dec-05 Mar-06Error    Figure 4. Difference between the sums of in-house circuits and of whole house meters (in %) -30.00%-25.00%-20.00%-15.00%-10.00%-5.00%0.00%5.00%10.00%15.00%20.00%Nov-04 Feb-05 May-05 Sep-05 Dec-05 Mar-06DayNight   Figure 5 Difference between the sums of in-house circuits and of whole house meters (day vs. night) -30.00%-25.00%-20.00%-15.00%-10.00%-5.00%0.00%5.00%10.00%15.00%20.00%Nov-04 Feb-05 May-05 Sep-05 Dec-05 Mar-06WEEDENDWEEKDAY   Figure 6. Difference between the sums of in-house circuits and of whole house meters (weekday /end) RESULT AND DISCUSSION   Aspen house electricity consumption analysis The consistency check between individual circuit and the whole house electricity meter indicates that the sensor readings are not of good quality. Not only is there significant difference between the sums of two types of (circuit vs. whole house) sensor readings, the difference (Fig. 4) varies with time and does not  Proceedings: Building Simulation 2007 - 1575 - show any recognizable pattern. When data is separated into day-night pair (Fig. 5) and weekday-weekend pair (Fig. 6) in monthly electricity load graphs, it is noted that besides PV system reading and life style of occupants, there are other unidentified factors contributing to this discrepancy y = 77.372e -0.5603x R 2  = 0.998105101520253035404550top5 top10 top15 top20 top25 sum of In-house circuits    U  n   i   t  :  p  e  r  c  e  n   t  a  g  e   (   %   ) sum of CTstrendline  Figure 7. Log scale (percentage of total electricity load of top ranked circuits) Living Room Befor May 2005 0.00000.20000.40000.60000.80001.00001.20001.40001.6000      0     0    :     0     0    :     0     0     0     1    :     3     0    :     0     0     0     3    :     0     0    :     0     0     0     4    :     3     0    :     0     0     0     6    :     0     0    :     0     0     0     7    :     3     0    :     0     0     0     9    :     0     0    :     0     0     1     0    :     3     0    :     0     0     1     2    :     0     0    :     0     0     1     3    :     3     0    :     0     0     1     5    :     0     0    :     0     0     1     6    :     3     0    :     0     0     1     8    :     0     0    :     0     0     1     9    :     3     0    :     0     0     2     1    :     0     0    :     0     0     2     2    :     3     0    :     0     0 time   a  v  e  r  a  g  e   A  m  p WeekdayWeekend   Living Room Starting May 2005 0.00000.10000.20000.30000.40000.50000.60000.70000.80000.9000      0     0    :     0     0    :     0     0     0     1    :     3     0    :     0     0     0     3    :     0     0    :     0     0     0     4    :     3     0    :     0     0     0     6    :     0     0    :     0     0     0     7    :     3     0    :     0     0     0     9    :     0     0    :     0     0     1     0    :     3     0    :     0     0     1     2    :     0     0    :     0     0     1     3    :     3     0    :     0     0     1     5    :     0     0    :     0     0     1     6    :     3     0    :     0     0     1     8    :     0     0    :     0     0     1     9    :     3     0    :     0     0     2     1    :     0     0    :     0     0     2     2    :     3     0    :     0     0 time   a  v  e  r  a  g  e   A  m  p WeekdayWeekend  Figure 8. Living room hourly electricity load before and after May 2005 On the circuit-by-circuit comparison, top 5, top 10, top 15, top 20 and top 25 circuits account for 46.0%, 70.7%, 84.4%, 92.7% and 97.5% of total electricity consumption respectively. There is a strong showing of “power law” (Fig. 7) relationship of electricity consumption among the top 25 circuits. The top 10 circuits are Living Room (16.9 %), Clothes Dryer (7.9 %), Air Handler (7.4 %), Kitchen Lights (7.0 %), Kitchen Refrigerator (6.8%), Maser Bedroom Power Plug and Lights (5.9 %), Garage Power Plug (5.2 %), Hot Water Boiler (5.2 %), Condenser (4.8%), and  North Bedroom Power Plug and Lights (3.7 %). Comparative graphs of two circuits are presented to demonstrate that the data quality problem does not affect the outcome of proposed solution as far as identifying opportunities to reduce electricity consumption is concerned. There were problems with readings of the living room (Fig. 8) plug-in load sensor and the sensor was “reset” by May 2005. The “before and after” graphs show that, as a result of this reset, the reading drop on average 0.1 Amp or 25% of the non-pick hours value. 25% error in reading is significant, but from the perspective of finding energy saving opportunity, it has little effect. Stand-by load, represented by the area under the non-pick hour line, still accounts for more than 50% of the total living room electricity consumption. With or without accounting for the 25% reading error, the proposed solution is still to eliminate the stand-by load.  Air Handler  0.00000.20000.40000.60000.80001.00001.2000      0     0    :     0     0    :     0     0     0     1    :     3     0    :     0     0     0     3    :     0     0    :     0     0     0     4    :     3     0    :     0     0     0     6    :     0     0    :     0     0     0     7    :     3     0    :     0     0     0     9    :     0     0    :     0     0     1     0    :     3     0    :     0     0     1     2    :     0     0    :     0     0     1     3    :     3     0    :     0     0     1     5    :     0     0    :     0     0     1     6    :     3     0    :     0     0     1     8    :     0     0    :     0     0     1     9    :     3     0    :     0     0     2     1    :     0     0    :     0     0     2     2    :     3     0    :     0     0 time   a  v  e  r  a  g  e   A  m  p WeekdayWeekend    Air Handler Monthly Average 0.00000.10000.20000.30000.40000.50000.60000.70000.80000.90001.0000      2     0     0     4     D     E     C     2     0     0     5     J     A     N     2     0     0     5     F     E     B     2     0     0     5     M     A     R     2     0     0     5     A     P     R     2     0     0     5     M     A     Y     2     0     0     5     J     U     N     2     0     0     5     J     U     L     2     0     0     5     A     U     G     2     0     0     5     S     E     P     2     0     0     5     O     C     T     2     0     0     5     N     O     V     2     0     0     5     D     E     C     2     0     0     6     J     A     N     2     0     0     6     F     E     B     2     0     0     6     M     A     R time   a  v  e  r  a  g  e   A  m  p  AVG(ahfahr)  Figure 9. Air handler hourly and monthly electricity loads The Air Handler graph (Fig. 9) reflects the life style of a typical American working family. To accommodate occupancy activities, AHU load pecks at early morning and late afternoon hours during weekdays, but pecks at morning and has lower peck in the afternoon during weekend. The monthly graph shows that the AHU electricity load pecks at summer and winter, which is understandable but it can not offer an explanation why 2005 winter in the W-  Proceedings: Building Simulation 2007 - 1576 - shaped curve consumed electricity twice as much as 2006 winter did. Although there is indication of  problems in the quality of the sensor reading, the hourly graph still provides useful information to help identify that we can reduce AHU electricity consumption by turning it off in weekday afternoon, when there is little occupancy activities. Master Bedroom 0.00000.10000.20000.30000.40000.50000.60000.70000.8000      0     0    :     0     0    :     0     0     0     1    :     3     0    :     0     0     0     3    :     0     0    :     0     0     0     4    :     3     0    :     0     0     0     6    :     0     0    :     0     0     0     7    :     3     0    :     0     0     0     9    :     0     0    :     0     0     1     0    :     3     0    :     0     0     1     2    :     0     0    :     0     0     1     3    :     3     0    :     0     0     1     5    :     0     0    :     0     0     1     6    :     3     0    :     0     0     1     8    :     0     0    :     0     0     1     9    :     3     0    :     0     0     2     1    :     0     0    :     0     0     2     2    :     3     0    :     0     0 time   a  v  e  r  a  g  e   A  m  p Weekdayweekend   Figure 10. Master Bedroom hourly electricity load Condensor  0.00000.10000.20000.30000.40000.50000.60000.7000      0     0    :     0     0    :     0     0     0     1    :     3     0    :     0     0     0     3    :     0     0    :     0     0     0     4    :     3     0    :     0     0     0     6    :     0     0    :     0     0     0     7    :     3     0    :     0     0     0     9    :     0     0    :     0     0     1     0    :     3     0    :     0     0     1     2    :     0     0    :     0     0     1     3    :     3     0    :     0     0     1     5    :     0     0    :     0     0     1     6    :     3     0    :     0     0     1     8    :     0     0    :     0     0     1     9    :     3     0    :     0     0     2     1    :     0     0    :     0     0     2     2    :     3     0    :     0     0 time   a  v  e  r  a  g  e   A  m  p WeekdayWeekend   Figure 11. Condenser hourly electricity load Kitchen Lights 0.00000.20000.40000.60000.80001.00001.2000      0     0    :     0     0    :     0     0     0     1    :     3     0    :     0     0     0     3    :     0     0    :     0     0     0     4    :     3     0    :     0     0     0     6    :     0     0    :     0     0     0     7    :     3     0    :     0     0     0     9    :     0     0    :     0     0     1     0    :     3     0    :     0     0     1     2    :     0     0    :     0     0     1     3    :     3     0    :     0     0     1     5    :     0     0    :     0     0     1     6    :     3     0    :     0     0     1     8    :     0     0    :     0     0     1     9    :     3     0    :     0     0     2     1    :     0     0    :     0     0     2     2    :     3     0    :     0     0 time   a  v  e  r  a  g  e   A  m  p WeekdayWeekend   Figure 12. Kitchen Lights hourly electricity load In complex system theory, functional units of different scales do not necessarily share the same  properties. A system which does not work globally does not mean its components do not work locally. Although sensor data is problematic in monthly scale, hourly data in the living room and AHU graphs vividly illustrate the usefulness of sensor data. Other graphs from top 10 circuits, such as master bedroom (Fig. 10), condenser (Fig. 11) and kitchen light (Fig. 12), also faithfully depict the life style of a typical American working family. Buffalo house energy conversion efficiency information model The case of Aspen house’s electricity consumption demonstrates what difference “scale” and “format” can make in the usability of sensor data. The concept of “local vs. global” view of sensor data is further extended to develop a complex system representation for the system performance of Buffalo house in terms of energy conversion efficiency. Two documents were acquired in the beginning: the system diagram of CHF/FC/Boiler marked with sensor points and the year-long sensor readings of these sensors. Because of the advantage of diagram over text on sharing of structured information (Larkin 1987), the system diagram (Fig. 13) is converted to a 3-tiered hierarchical structure, also presented in a diagrammatic form (Fig. 14). Top tier view (A0) shows the energy conversion efficiency from aggregate effects of all components in the system. Second tier views show the conversion efficiency of two sub-systems – the CHP/FC system (A1) and conventional boiler system (A2). Third tier views show the energy conversion performance of four fundamental product units belong to those two sub-systems- FC chemical energy reaction unit (A11) , FC exhaust heat recovery unit (A12) (components of CHP/FC system) and domestic hot water (DHW) heat exchange unit (A21), space heating heat exchange unit (A22) ( components of boiler heat exchange system). A main characteristic of this hierarchical information model is that it maps the multi-resolution views to actual product units. The knowledge of problem solving is thus embedded in this system because the information model not only detects problems but also pinpoint the problematic component thus the corresponding specialist can be called in to deal with it. An example of the home-owner self-help continuous commissioning is illustrated as following. The focus of this example is the diagnostics of the A22 space heating boiler. We pick 3 days (152, 302, 362) in year 2004 as inputs. After submitting the times, the information model (Fig. 14) generates with a table of energy conversion efficiency of various  process units and the detailed efficiency graph of A22 process unit (the boiler for space heating). On day 152 (Fig. 15), the resulting table indicates  problem with the space heating boiler. The detailed graph shows that there is no energy inflow and outflow and water temperatures are constant. Thus the diagnostics is that the boiler is not in operation. It is reasonable considering that day 152 is not in the heating season. On day 302 (Fig. 16), again the resulting table indicates problem with the space heating boiler. The A22 graph shows that there is energy outflow but no energy inflow and water flow
Advertisement
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks