International Journal of Computer Science and Applications © Technomathematics Research Foundation Vol. 11 No. 1, pp. 1 - 17, 2014

DATA ANALYSIS IN THE INTELLIGENT BUILDING ENVIRONMENT

DALIA KRIKSCIUNIENE Faculty of Informatics, Masaryk University, Brno, Czech Republic, [email protected] TOMAS PITNER Faculty of Informatics, Masaryk University, Brno, Czech Republic, [email protected] ADAM KUCERA Faculty of Informatics, Masaryk University, Brno, Czech Republic, [email protected] VIRGILIJUS SAKALAUSKAS Department of Informatics, Vilnius University, Vilnius, Lithuania, [email protected]

The article addresses the problem of intelligent analysis and evaluation of facility management data gathered from heterogeneous sources, including environmental data collected from building automation sensors, temporal weather characteristics and scheduling. We suggest the framework of analytical model, based on deriving descriptors which could sentinel the level of thermal comfort of working environments. The model aims to facilitate process of extracting essential characteristics of facility management for detecting dependencies and observing anomalies. The framework aims to discover hidden relations between performance of building conditioning and environmental and spatial factors that cannot be observed from the building automation system itself. The performance of the model was tested by experimental analysis of facility management of the university cam- pus, designed for exploring how various environment variables affect temperature in the lecture rooms, equipped by the air conditioning devices. Based on the obtained results, we elaborate on further steps needed for beneficial, efficient and flexible data analysis system for the field of facility management. Keywords: Facility management; computational intelligence; machine learning; sensor networks; system integration; environmental conditions. 1

2

Kriksciuniene, Pitner, Kucera, Sakalauskas

1. Introduction Modern buildings tend to be equipped by various technologies, thus raising the problem of facility management. IFMA ( International Facility Management Association) defines facility management as ”a profession that encompasses multiple disciplines to ensure functionality of the built environment by integrating people, place, process and technology” [IFMA (2014)]. The complexity of the high-level decision-making in this area emanates from lack of knowledge and undefined causal relationships between data flows coming from multiple sources including low-level sensor networks, geographical data systems (GIS), weather conditions and various other inputs taken into account by facility managers according to their experience and expertise. The concept of “intelligent buildings” is becoming standard for monitoring building sites, where the prevailing approach is based on cost and investment analysis [Wong, Li and Wang (2005)]. This domain area can also be analyzed as interplay of related subsystems where each of them follows its own goal, such as maximizing perceived comfort by the users, increasing quality and effectiveness of facility control procedures, providing support and advice for facility management personnel, at the same time minimizing energy consumptions and maintenance costs. The problem which emanates from the perspective of system interplay can be formulated as search for methods which can resolve the gap between indicators placed in different time scales: the cost and energy consumption indicators used for control are “lagging”, as they reflect past performance. The data provided by sensors and other environmental characteristics can be called “leading”, as they contain hidden information advancing future value of the “lagging” indicators. The problem is also affected by the “intermediate” variables describing comfort of the users which have individual tolerance ranges to changing conditions of the building environment. In many cases the facility managers are not supplied by the information of the “lagging” variables, as their decisions have to be based on real-time data flow. In this article we focus on intelligent analysis of building data by searching causal interrelationships between “leading” variables which could lead to deviations and anomalous states of facility management system. Sections 1 and 2 describes domain of facility management, presents the overview of state-of-the-art in the use of computational intelligence and machine learning methods in the fields of automation and regulation of intelligent building operation based on monitoring sensor network. Section 3 proposes the method for building management data analysis and describes use case we performed computations on. Section 4 evaluates result of the proposed method. Section 5 elaborates on further steps toward complex tools for intelligent facility management data analysis. The last section concludes the results and outlines topics for further research.

Data Analysis in The Intelligent Building Environment

3

2. Intelligent Facility Management Systems Construction and operation of modern building or multi-building site is the interdisciplinary research area, where various scientific fields and application domains are interrelated (Fig. 1).

Fig. 1. Operation of modern facility management.

All the aspects of building operations are closely connected for building environment-friendly and cost effective buildings, able to provide appropriate environment conditions for their inhabitants as well [Prez-Lombard, Ortiz and Pout (2008)]. This task is achieved by improving control loops in regulation and automation with the help of various techniques such as machine learning or modeling applied for various types of data sets. Operation of the building system is monitored and controlled by using data provided by sensor networks, operator workstations, web applications, tenant portals, and archive databases. The complexity levels of analysis include data mining algorithms, complex processes and event analysis. Facility managers need tools which simplify management and maintenance by integrating data from various sources, allowing remote control of different technologies from one access point. In large organizations these systems may include services such as helpdesk, preventive maintenance and inspection evidence, room reservations, space planning or tools for real-time controlling and monitoring of various building systems [Atthajariyakul and Leephakpreeda (2004)]. Various standard systems for security, maintenance or access control of building, such as CCTV (closed circuit television), lighting control, elevator monitoring, and HVAC (heating, ventilating, air conditioning) are integrated into complex BMS

4

Kriksciuniene, Pitner, Kucera, Sakalauskas

(Building Management System) system for monitoring and control of all subsystems installed in the particular building (Fig. 2).

Fig. 2. Building management system, BMS.

The BMS provides bus (usually computer network) designed for merging data from various systems by using common communication protocol (BACnet, KNX/EIB, LON) [Merz, Hansemann and Hbner (2009)]. The architecture of BMS consists of several operator workstations, web server, archive server connected to this bus, which allows monitoring and controlling current state of systems in the building. The research literature focusses on analysis, optimization, device scheduling based on the historical data provided by various sensors (temperature, humidity, energy consumption sensors) and performing computations based on historical trends observed in the data time series. A number of intelligent systems can be used for these tasks [Doukas, et al. (2007)], such as application of artificial neural networks [Atthajariyakul and Leephakpreeda (2008)], [Yang and Kim (2004)], [Moon and Kim (2010)], [Mohanraj, Jayaraj and Muraleedharan (2012)], decision support [Shen, Hao and Xue (2012)], expert systems [Orosa (2011)] fuzzy logic control systems [Dounis and Manolakis (2001)], [Kristl, et al. (2008)], simulation, clustering or outlier detection [Seem (2007)], [Seem (2005)]. The selection of methods highly depends on data sources. The most common indicators for building operation efficiency evaluation are based on the “lagging” indicators, such as overall building electric energy consumption. These data can be obtained by connecting main electricity consumption meter to the Building Automation System (BAS) with the help of specialized protocols

Data Analysis in The Intelligent Building Environment

5

such as M-BUS or MODBUS; or higher level protocol (such as BACnet) [Seem (2007)], [Seem (2005)], [Xiaoli, Bowers and Schnier (2010)]. The delay of providing consumption information creates main drawback of existing analytical models and is often solved by using simulated data. Electricity consumption data can be obtained not only by providing absolute values, but also as daily increments, which serve as useful indicator for comparative consumption analysis of buildings with similar technical equipment. Another type of metrics of HVAC is user centered. They include number of user complaints related to the ambient conditions of the building or Predicted mean vote (PMV) index. This index estimates number of discomforted users and integrates various parameters such as clothing, characteristic of human metabolism and various environment variables ( temperature, humidity). The PMV index is suitable as optimizing parameter for HVAC system, or regulation of the heating units [Cigler, et al. (2012a)], [Cigler, et al. (2012b)]. As the user centered approach explores ranges of variables defining comfort of environmental conditions, we can assume that avoiding outliers and anomalous values can lead to optimal values of the energy consumption and costs. The anomalous processes often mark fault or wrong setup of the building automation or inappropriate use of the building equipment [Seem (2007)], [Seem (2005)], [Xiaoli, Bowers and Schnier (2010)]. In this article we focus on analysis for securing thermal comfort of building environment reflected by sensor data of HVAC systems. As HVAC operation expenses make significant share of overall building operation cost, therefore efforts to make HVAC system more efficient are highly feasible. 3. Intelligent HVAC Analysis Method We suggest method for evaluating impact of factors which could influence room environment and in particular the stability of air temperature. We utilize intelligent analysis methods allowing facility managers to compare combined effect of available factors which tend to significantly vary over time, and to decide which of them should be taken into account during building operation optimization. For experimental analysis we focus on the use of easily accessible data of room temperatures, supplied from the archive server and joining data from several other sources, in order to extract easily understandable characteristics of complex processes which affect HVAC operation. The method was previously presented at IIMSS 2013 conference and published in [Kriksciuniene, et al. (2013)]. The novel feature of the model is based on intelligent analysis of building data by exploiting causal interrelationships between the derived indicators and arranged in the four levels based on their cause-effect relationships, in order to resolve the gap between indicators placed in different time scales: “lagging” and “leading”. The indicators of cost and energy consumption are desired to be used for reflecting past performance and control. However these indicators are not available at the moment

6

Kriksciuniene, Pitner, Kucera, Sakalauskas

of decisions of facility managers. Therefore the real- time data provided by sensors and other environmental characteristics can be called “leading”, as they contain hidden information capable to forecast future value of the “lagging” indicators of resource consumption and serve as basis for decision making. The analytical model consists of two compounds: (1) Preparing data flow for analysis. It consists of deriving “leading” indicators by extracting and aggregating data from internal and external systems potentially having the causal relationships and capacity of influencing “lagging” indicators. In Fig. 3 the suggested levels of indicators are presented: measures included to the level of “Resource indicators” represent the “lagging” indicators which are influenced by the “leading” indicators presented in the levels of “User centered metrics”, “Building parameters” based on construction qualities and also by the effects of “Weather conditions”. The values of the selected indicators can be distorted or affected by the behavior of users of the facilities. In order to reveal these effects the level of “Dwelling processes and habits” is formed of indicators describing choices made by users and influence of number of people present in the facilities. (2) Applying computational intelligence methods for defining causal relationships between levels of indicators and also for defining most influential indicators for analysis. In the presented analysis we applied clustering, regression, sensitivity analysis, which can further lead to application of hierarchical fuzzy and anomaly detection methods. 3.1. Context The experimental use case for illustrating the depth of the problem domain is selected on the basis of Building Management System (BMS) used for controlling and monitoring operation of Masaryk University campus at Brno, Czech Republic. The construction of the campus started in 2007. The concept of connecting all the facility management data by capturing systems of over 25 buildings into the BMS was implemented by using BACnet protocol. The network consists of two web servers, two archive servers, operator workstations, and over 700 devices. These devices function as PLCs ( programmable logic controllers) for controlling mainly HVAC systems and application gateways for the connected systems of access control, security system and fire safety system. This enormous data flow is interconnected with the data from other systems of Masaryk University, such as GIS database of building passport, or Academic information system. The detailed data flow builds basis for designing advanced analysis of BMS. 3.2. Setup The experimental data is composed of inputs from 15 lecture rooms which belong to one of the buildings of University campus. Each room is equipped with local (in-

Data Analysis in The Intelligent Building Environment

7

Fig. 3. Levels of indicators included to analytical model.

room) air-conditioning (AC) units, which can be controlled by control panel located inside the room. Users are enabled to switch AC unit on or off and change desired room temperature. AC unit controls the speed of fan which supplies cold air into the room, opens or closes valves of central heating radiators. AC unit automatically switches off if windows in the room are opened. The examined rooms are located over all three floors of the building; they also differ in size, area, geographical orientation and number of the windows. These data are stored in spatial database called Building passport of Masaryk University, which contains geographical data of all the buildings of the university. 3.3. Measurement The total size of dataset is 61920 records. It involves 4128 records per each of 15 lecture rooms. We assume that the thermal regime is optimal if the room temperature has high stability over time series. For the preliminary exploring of the use case we choose the output characteristics of standard deviation of inside temperature in relationship to 22 ◦ C as prevailing preset temperature of the conditioning of lecture rooms. The investigation is applied for the data registered between 6:00 and 20:00. In this period room environment is infuenced by number of factors that are not present in the night (number of occupants, amount of sunlight). The time period of analysis is from 1st of April to 13th of May suitable to

8

Kriksciuniene, Pitner, Kucera, Sakalauskas

observe how weather conditions can affect environment inside the buildings without influence of heating system. Data about actual and desired room temperature were obtained by using 15 minutes polling. Weather conditions (outside temperature, pressure, humidity, downfalls) were obtained from the on-line free meteorological information source [InMeteo (2012)] for Brno location with 30 minutes polling. We joined the time series of room temperature from BMS together adequately with data from the meteorological station. 3.4. Processing After processing data by applying intelligent analysis module of Statsoft tool [StatSoft Inc. (2011)] the seven aggregated data variables are presented in Table 1: standard deviation temperature (Stn dev), inside temperature mean (T I), height (H), number of windows (W), room area (A, m2 ), geographical orientation O (East(E), West(W), inside (I) of building), and maximum capacity for people in room (C).

Table 1. Aggregated variable data.

Room No. BHA12N03034 BHA12N02006 BHA12N01032 BHA12N02035 BHA12N02036 BHA12N02005 BHA12N02034 BHA12N03027 BHA12N03005 BHA12N03011 BHA12N01014 BHA12N03033 BHA12N02011 BHA12N03006 BHA12N02028

Stn dev 0.90 0.91 1.04 1.06 1.08 1.13 1.24 1.31 1.39 1.51 1.59 1.92 2.15 2.67 2.77

TI 22.00 22.43 22.20 22.58 22.66 22.73 22.50 22.97 22.61 23.17 21.33 23.56 24.01 24.32 24.60

H 2.80 2.80 4.15 2.80 2.80 2.80 2.80 2.80 2.80 2.80 4.15 2.80 2.80 2.80 2.80

W 2 13 0 2 2 13 13 12 13 12 0 7 12 13 12

A 164 116 217 58 56 126 126 84 126 84 217 79 67 114 68

O W E W W I I E E E W W I I E E

C 259 237 126 114 40 40 32 50 50 126 114 40 40 32 130

The first step of analysis aimed to group rooms by their characteristics. In Table 1 the lecture rooms are ranked according to the value of standard deviation. Three groups can be distinguished: smallest values of Stn dev are registered for the four rooms with large capacity and area, the medium Stn dev group contains five rooms with small capacity and the last group with the largest values of Std. is mixed in all characteristics, which may imply differences in technical equipment installed

Data Analysis in The Intelligent Building Environment

9

or other root causes not reflected by the collected data. The average temperature T I tends to increase together with the increasing value of Stn dev, it suggests low cooling capacity, higher influence of outer weather characteristics or ineffective facility management. At the second step, the values of standard deviation are explored in relationship to the room characteristics and the weather conditions. In Table 1 we observe the strongest positive relationship of Stn dev to height and number of windows; the values of Stn dev tend to decrease if the area of room increases. The three characteristics of outer weather conditions do not affect Stn dev according to values of p-level in Table 2.

Table 2. Summary of multiple regression for full data set.

From Table 2 we assume that the importance of the variables changes in all three groups of rooms (Table 1). As the 3rd group consists of rooms with almost identical characteristics as in group 1, we can assume that the main influence lies outside the supplied data for analysis variables. E.g. the rooms BHA12N03011, BHA12N01014 and BHA12N02028 have high capacity but the conditioning quality is low. In the following steps we analyzed the influence of variables for each separate group of rooms. In each group the weather conditions (outer temperature, humidity, and pressure) had no influence for the Stn dev values. In Table 3 the results are combined for the two groups of medium and low values of standard deviation (Stn dev). Here the construction-based characteristics of rooms have even bigger influence, as is expressed by the increase of p-level. At the third step, the data set was explored by cluster analysis in order to get insight of distribution of values of the variables affecting stability of thermal comfort. The Viscovery Vsomine tool [Viscovery Vsomine (2008)] was applied for data analysis. The best separation with different interrelated effect of the variables was achieved by making four clusters. In Fig. 4 and Table 4 can observe big differences in ranges of the variable values and the consequent values of the Stn dev (both negative and positive). Each cluster has its distinguished characteristic which can be assumed as a separation basis

10

Kriksciuniene, Pitner, Kucera, Sakalauskas

Table 3. Results of subset of medium and high values Standard deviation (Stn dev).

Fig. 4. Variable value distribution of among clusters.

(large height and absence of windows for C4, lowest value of Stn dev for C3).

Data Analysis in The Intelligent Building Environment

11

Table 4. Clusters characteristics.

3.5. Interpretation of results The experimental research showed that the enormous technical data flow does not provide sufficiently meaningful information for decision making in facility management. The root causes lay in the influences which are not properly captured by sensor data or intuitively overestimated (as weather conditions in our research). The further research of revealing important “leading” variables is planned in the direction of measuring influence of time of people presence in rooms, space volume per person, etc. and also refinement of the characteristics related to registering operating periods of the conditioning devices and their preset conditions. The design of workflow for clustering analysis showed that cluster separation implies combined influence of varying ranges for the variables, which suggests further use of fuzzy methods for investigation.

4. Model Evaluation We use the computational approach of analysis, which assumes that the analytical methods have to explore combined influence of numerous factors from inside and outside environments in order to reveal their influence and sensitivity of decision making by facility managers. The novel feature of the model is based on intelligent analysis of facility management data by exploiting causal interrelationships between the derived indicators which are designed from the sensor based facility management system and arranged in four different type of levels based on their cause-effect relationships, in order to resolve the gap between indicators placed in different time scales: “lagging” and “leading”. The experimental research is based on analysis of sensor network data flow of university campus. The research showed that the indicator of stability of thermal comfort has different dependence on the analyzed variables. The outer variables of weather temperature, humidity and pressure had lowest influence, the room height, number of windows and square area had different importance in separate clusters having low, medium and high values of thermal deviation. However application of temperature sensors for facility management decision making was not sufficient, therefore, increasing number of variables and exploring their importance for the thermal comfort can increase precision of analysis.

12

Kriksciuniene, Pitner, Kucera, Sakalauskas

5. Towards Intelligent Facility Management The problem of the research in the area of facility management is highly influenced by the interdisciplinary characteristics: it depends on construction technologies, intelligence and quality of control devices, and approaches for analysis. The problem which emanates from the perspective of system interplay can be formulated as search for methods which can resolve the gap between indicators placed in different time scales: “lagging” indicators of cost and energy consumption are used for reflecting past performance and control. The data provided by sensors and other environmental characteristics can be called “leading”, as they contain hidden information advancing future value of the “lagging” indicators. As decisions of facility managers have to be based on real-time data flow, our research setting was designed to rely only on “leading” characteristics. In this article we focus on intelligent analysis of building data by searching causal interrelationships between variables which could reveal deviations and anomalous states of facility management systems. 5.1. Current state Data analysis based on “lagging” indicators is sufficiently supported in CAFM (Computer-Aided Facility Management) software systems, but such tools don’t provide wide support for analysis of building operation data. Both CAFM products and building management systems are supposed to help the facility managers to reduce building operational cost. Interconnection and cooperation of those separate systems is a promising way to optimize fault detection and recovery and shorten the delay between acquiring the data by the BMS and performing actions leading to optimization of building operation. One of the possible approaches to integration of various information systems used by facility managers is described in [Shen, Hao and Xue (2012)]. Next, we describe proposed integration methods for BMS based on BACnet protocol and Archibus CAFM software in the environment of the UKB. We will discuss two sample integration use cases - simplifying maintenance workflow and energy consumption analysis. 5.2. Building maintenance On-Demand maintenance support in the CAFM software relies on system of requests. A request is usually entered into the system by maintenance staff or facility users. Then, it passes different phases (assigning, scheduling, material and tool requesting, billing, checking) of its life cycle which ends when the issue is fixed. This process of fault resolution could be simplified by automated approach to creating requests in the CAFM software whenever fault is detected. For example, rooms with unsatisfactory temperature comfort detected by tool presented above could produce request for maintenance of AC units in affected rooms. The main problem of this approach is caused by insufficient data validity. For the on demand maintenance system to be effective, the CAFM software must not

Data Analysis in The Intelligent Building Environment

13

be polluted by false alarms, trivial, or temporary problems. Complex algorithms implemented in advanced analytical tools can help with the task of filtering valid issues that should be transferred into the CAFM system. 5.3. From lagging to leading Another example use case of integration combines CAFM software ability to analyze “lagging” data and uses the same approach for on-line “leading” indicators coming form BMS. The energy management in the CAFM software utilizes data from energy vendor invoices. This data is available only for the building as a whole because the energy meter used for accounting is common for the building or even for the whole site. The BMS however allows to gather much more detailed data about the energy consumption (the electricity consumption in particular) to the level of individual devices such as AC drives (variable-frequency drives used for control of AC motors of air conditioning units). Furthermore, the consumption data are far more recent than consumption data received form the energy vendor once a month. The precise data coming from the BMS together with analytical capabilities of the CAFM tools can significantly improve the insight into the energy efficiency of buildings brought to the level of individual devices, systems and rooms. To make this seamless cooperation of two disciplines possible and bring similar easy-to-use tools as those available for the “lagging” indicator analysis to the field of “leading” data analysis, further steps in system integration must be taken. 5.4. Towards more semantics Currently available data of the BMS at Masaryk University do not contain information about their meaning by default. Each data point is identified only by its network address according to BACnet protocol specification. Name of the data point is designed to be understandable by human operators and is unsuitable for machine processing. Semantic information such as location of the sensor or measured quantity must be added manually for the purposes of analysis. This drawback makes data analysis inflexible and prolongs the data analysis workflow. Ad-hoc approach to system integration is sufficient for experimental purposes such as testing new methods of data analysis and provides valuable results concerning evaluation of different approaches, but prevents deployment of proposed methods for routine operation. Complexity of the system and lack of semantic information about gathered data in fact prevents facility management staff to perform data analysis routinely and on regular basis. The necessity of ad-hoc approach to linking data from various sources also hides complex relations between data points, devices, environmental variables and other factors that influence building operation. As mentioned in the evaluation of experimental part of this article, to fully understand room temperature change patterns, we need to utilize large number of other indicators that are unavailable in current

14

Kriksciuniene, Pitner, Kucera, Sakalauskas

setup. Some of them are not measured at all, some of them are not stored in archival database or they are available in information system that is not freely accessible. On the other hand, some additional data about building operation influencing temperature in examined rooms are gathered and stored by BMS. Even though, size and complexity of the system hides certain relations between operation data from various sources which limits the human operators in meaningful usage of such data in the task of post-mortem operation analysis. Typical example of such hidden relation is the link between state of pumps, motors and valves of central heating unit placed in the utility room in the basement. Heating unit is monitored and controlled by different part of the BMS/BAS than the local AC unit in the lecture room. The local AC controller has ability to control the valve on the heating radiator in the room. Actions of the local AC controller are thus influenced by current setup of central heating unit (if the central heating is off, controlling of the valve of the room radiator does not influence room temperature at all). The relation of central heating unit setup and operation of local AC unit cannot be observed from the archive data itself. Some information about complex relations can be derived by examining regulation algorithms and detecting used variables, input and outputs, or data points. Other relations are not described in the BMS or BAS itself at all because it is determined by physical installation of various devices in the building. In the case of multiple central heating units, BMS/BAS does not contain information concerning physical plumbing - it is impossible to determine which radiator valve is connected to which central heating unit. 5.5. Ontologies for intelligent buildings In order to make easy and fast retrieval of all the related data from the BMS, ontology model of relations between different elements in the building (devices, locations, data points, affected variables, measured physical quantities, . . . ) is needed. Creation of such model is made possible by existence of BIM (Building Infrastructure Modeling) data sources containing information about facilities of Masaryk university and physical installation and connection of all elements of the building equipment. In case of Masaryk University, we are talking about spatial databases of “building passport” and “technology passport”. Extraction of information about physical connection between devices allows us to create complete and complex model of building environment. Construction of the model is challenging task that combines automated and semi-automated procedures (analysis of regulation algorithms, parsing electronic building documentation, extraction data from BIM sources) with expert work of facility management stuff (definition of the model, development of automation tools, resolving conflicts, gathering BIM data). 5.6. Future applications With the model available, advanced data analysis applications can be developed. Querying the database of system model will enable applications to gather informa-

Data Analysis in The Intelligent Building Environment

15

Fig. 5. Schema of FM systems integration.

tion from all the data points containing data related to the current object of interest. Figure 5 presents scheme of BMS system enriched by additional data sources and advanced applications. BMS serves as source of building operation data. Archive server is relational database optimized for transaction processing (OLTP). Archive data are transferred to the data mart which serves as OLAP data store for business intelligence applications. Data in the data mart are enriched by semantic information gathered from spatial databases of BIM (technology passport). Data from data mart can be also used by CAFM software tools in the same way as they use “lagging” data. BMS can additionally serve as source of another type of information besides archived data from the database. Events generated by the elements of the BMS and on-line sensor data are obtained by monitoring applications and Complex Event Processing engine. Those tools serve for detection of faults or anomalies and outliers in the system behavior, allowing BMS operators to identify faulty parts of the system. 6. Conclusions This article introduces interdisciplinary field of facility management and presents contrast between “lagging” performance indicators such as data obtained from energy vendors and “leading” indicators obtained directly from various building systems integrated in BMS (Building Management System). While analysis of “lagging” indicators is well supported in current state-of-the-art CAFM (Computer-

16

Kriksciuniene, Pitner, Kucera, Sakalauskas

Aided Facility Management) software, analysis of “leading” indicators becomes challenging research topic with the spread of intelligent buildings. Article proposes model and method for evaluation of room comfort based on data obtained from heterogenous systems and data sources. Data from BMS, room reservation system and weather station data provider were joined and resulting data sets were processed using statistical and computational knowledge methods. The approach combines data obtained fro building automation system with other factors that are not taken into account in regulation algorithms. Application of machine learning methods aims to reveal complex relations between performance of building conditioning systems and evironmental and spatial factors such as weather conditions and room orientation or number of windows. Next, we elaborate on future work in the field of building operation data based on the results provided by experimental analysis of room comfort in multiple lecture rooms at University campus of Masaryk University. We propose development of ontology model of building systems and devices relation that will be able to speed up the analysis workflow (selecting appropriate data, joining inputs from various sources, processing by analytical engine) and provide new flexible and easy to use applications and tools for facility management in large organizations. References Atthajariyakul, S.; Leephakpreeda, T. (2004): Real-time determination of optimal indoor-air condition for thermal comfort, air quality and efficient energy usage, Energy and Buildings, Volume 36, Issue 7, Pages 720-733, ISSN 0378-7788, 10.1016/j.enbuild.2004.01.017. Atthajariyakul, S.; Leephakpreeda, T. (2008): Neural computing thermal comfort index for HVAC systems, Energy Conversion and Management, Volume 46, Issues 1516, September 2005, Pages 2553-2565. Cigler, J., et al. (2012): Optimization of Predicted Mean Vote index within Model Predictive Control framework: Computationally tractable solution, Energy and Buildings, Volume 52, Pages 39-49, ISSN 0378-7788, 10.1016/j.enbuild.2012.05.022. Cigler, J., et al. (2012): On Predicted Mean Vote optimization in building climate control. 2012 20th Mediteranian conference on control & automation (MED). Barcelona, Spain. Pages 1518-1523, ISBN 978-1-4673-2531-8/12/2012IEEE. Doukas, H., et al. (2007): Intelligent building energy management system using rule sets, Building and Environment, Volume 42, Issue 10, Pages 3562-3569, ISSN 0360-1323, 10.1016/j.buildenv.2006.10.024. Dounis, A. I.; Manolakis, D. E. (2001): Design of a fuzzy system for living space thermalcomfort regulation, Applied Energy, Volume 69, Issue 2, Pages 119-144, ISSN 03062619, 10.1016/S0306-2619(00)00065-9. InMeteo, s.r.o. available at: http://www.in-pocasi.cz/meteostanice/stanice.php?stanice=brno referred on June 01, 2012 International facility management association (2014): What is facility management?, available at: http://ifma.org/about/what-is-facility-management, referred on January 09, 2014. http://www.ifma.org/ Kriksciuniene, D., et al. (2013): Sensor Network Analytics for Intelligent Facility Manage-

Data Analysis in The Intelligent Building Environment

17

ment, Frontiers in Artificial Intelligence and Applications, Volume 254, Pages 212 221, ISSN 0922-6389, 10.3233/978-1-61499-262-2-212 ˇ et al. (2008): Fuzzy control system for thermal and visual comfort in Kristl, Z., building, Renewable Energy, Volume 33, Issue 4, Pages 694-702, ISSN 0960-1481, 10.1016/j.renene.2007.03.020. Merz, H.; Hansemann, T.; Hbner, C. (2009). Building Automation: Communication systems with EIB/KNX, LON and BACnet (Signals and Communication Technology). Springer-Verlag: Berlin, Heidelberg. ISBN-13: 978-3540888284 Mohanraj, M. ; Jayaraj, S.; Muraleedharan C. (2012): Applications of artificial neural networks for refrigeration, air-conditioning and heat pump systemsA review, Renewable and Sustainable Energy Reviews, Volume 16, Issue 2, Pages 1340-1358, ISSN 13640321, 10.1016/j.rser.2011.10.015. Moon, J. W.; Kim, J. (2010): ANN-based thermal control models for residential buildings, Building and Environment, Volume 45, Issue 7, Pages 1612-1625, ISSN 0360-1323, 10.1016/j.buildenv.2010.01.009. Orosa, J. A. (2011): A new modelling methodology to control HVAC systems, Expert Systems with Applications, Volume 38, Issue 4, Pages 4505-4513, ISSN 0957-4174, 10.1016/j.eswa.2010.09.124. Prez-Lombard, L.; Ortiz, J.; Pout, C. (2008): A review on buildings energy consumption information, Energy and Buildings, Volume 40, Issue 3, Pages 394-398. Seem, J. E. (2007): Using intelligent data analysis to detect abnormal energy consumption in buildings, Energy and Buildings, Volume 39, Issue 1, Pages 52-58, ISSN 0378-7788, 10.1016/j.enbuild.2006.03.033. Seem, J. E. (2005): Pattern recognition algorithm for determining days of the week with similar energy consumption profiles, Energy and Buildings, Volume 37, Issue 2, Pages 127-139, ISSN 0378-7788, 10.1016/j.enbuild.2004.04.004. Shen, W.; Hao, Q.; Xue, Y. (2012): A loosely coupled system integration approach for decision support in facility management and maintenance, Automation in Construction, Volume 25, Pages 41-48, ISSN 0926-5805, 10.1016/j.autcon.2012.04.003. StatSoft Inc. (2011): Electronic Statistics Textbook. Tulsa, OK:StatSoft, available at: http://www.statsoft.com/textbook/ stathome.html, referred on June 01, 2012 Viscovery Vsomine (2012): Explorative data mining based on SOMs and statistics, available at: http://www.viscovery.net/somine/, referred on June 01, 2012 Wong, K.W.; Li, H.; Wang, S.W. (2005): Intelligent building research: a review, Automation in Construction, Volume 14, Issue 1, Pages 143-159, ISSN 0926-5805, 10.1016/j.autcon.2004.06.001. Xiaoli, L.; Bowers, C. P.; Schnier, T. (2010): Classification of Energy Consumption in Buildings With Outlier Detection. IEEE Transactions On Industrial Electronics 57, no. 11: pp. 3639-3644. Yang, I.; Kim, K. (2004): Prediction of the time of room air temperature descending for heating systems in buildings, Building and Environment, Volume 39, Issue 1, Pages 19-29.