This post illustrates how to exploit data already collected within commercial buildings for energy management applications.
Data is key to unlocking energy efficiency savings in commercial buildings. Yet, most energy management solutions cannot gather the amount and heterogeneity of data required while maintaining a reasonable payback period. One opportunity often neglected is whether buildings’ existing infrastructure could, in lieu, be exploited to retrieve streams of data, thereby precluding costly measurement devices and deployment burden.
Data is ubiquitous in commercial buildings, and although the data’s original data collection purpose may be irrelevant to energy use, its production and sole existence may intimately be related to an event that connects to a specific energy event. For instance, most data packets flowing across a LAN are transmitted from PC-class machines; as such, the capture of a network packet tells that a machine is powered on, and that a person might be using that machine and therefore be physically present within the building at the time the packet was transmitted. Using existing legacy data sources for a purpose other than for which it was originally provisioned is an opportunity that has as yet not been explored sufficiently.
This post presents primary findings gathered from a thorough exploration of data sources available in commercial buildings. Data sources with the potential to uncover a wide range of energy insights have been identified and are analysed. The impediments inherent to infrastructure reuse are also discussed.
1. Data Networks 1.1 Virtual local area networks 1.2 Sensor networks 1.3 Telephone network management systems 2. Electrical Installation 3. Miscellaneous Systems 3.1 Machine registration 3.2 LDAP access logs 3.3 Equipment setback timing strategies 3.4 Swipe cards 3.5 Work schedules 3.6 Third party data 4. The Challenge of Integrating Data Sources 4.1 Open systems 4.2 Closed systems 5. Conclusions
1. DATA NETWORKS
Data networks are communication infrastructures used to send and receive data in an organised way. Local area networks (LANs), sensor networks and telephone networks are data networks widely used in office buildings. Many organisations active with green networking/computing have focused on reducing the energy usage of the computing infrastructure itself, from the development of energy-efficient computing equipment [1] to the introduction of low-power communication protocols [2]. Data networks as a framework for opportunistic electricity monitoring has nevertheless not be researched to a great extent.
>Mechanisms of office buildings’ data networks are discussed next to highlight how judicious data exploitation enables energy insights to be gathered.
1.1 Virtual local area networks (VLANs)
Original purpose
Local area networks (LANs) are computer networks covering small physical areas, generally using Ethernet over twisted pair cabling and Wi-Fi as data-transfer technologies. Virtual local area networks (VLANs) add logical grouping of network nodes, in order to reduce broadcast traffic and facilitate scalability, security, and network management. Logical subnets in the form of VLANs overlay the physical LAN, meaning that a computer at any location on the physical network can participate in a VLAN. Therefore, users in different buildings can be part of the same VLAN group, broadcasting messages to one another and sharing the same group servers, printers, and other resources.
Monitoring a VLAN helps network administrators for network inventory and for management of networked equipment. VLAN monitoring tools generally make use of Internet Protocol (IP) and transport layer packets to discover connected hosts, security holes and other network performance indicators. A typical method to discover hosts on the network is to send Internet Control Message Protocol (ICMP) Echo requests, which should prompt target hosts to respond with ICMP Echo reply messages [3]. However, ICMP packets are often filtered out and alternate techniques may be used to learn about connected hosts, e.g. Address Resolution Protocol (ARP) Pings and Dynamic Host Control Protocol (DHCP) cache access. ARP Ethernet MAC-to-IP address entries are stored within an ARP cache on routers during the time they are being used. With ARP, cache entries are removed when no packets are sent for a given period. A DHCP server also maintains a table of which addresses are currently leased-out. With DHCP, hosts must renew DHCP leases within a given period. The DHCP server will return the IP address to the pool of free IP addresses when a machine is powered off or disconnected from the network. Turnkey VLAN monitoring tools employing all possible communication protocols are able to return more machine information such as machines’ operating system, running services and network cards’ vendor information. Data traffic and content analysis is also targeted by network administrators to uncover illegal network activity, e.g. virus and downloads.
Evidence of electricity spending
Norman et al. [4] discussed in a short abstract the principle of using existing networks for energy purposes. They introduced the notion of implicit occupancy sensing, where localised space occupancy profiling is obtained using data from existing IT networks. In-depth techniques and real-world experimentation of implicit occupancy sensing and its potential for energy applications has however not been conducted until recently [5, 6, 7]. Consider that ARP and DHCP cache both indicate the current number of active hosts within a local network and therefore the number of machines powered on within a building at a given time. Using ARP and DHCP tools therefore offers a powerful opportunity to track the current power state of networked equipment, and to assess the power contribution and usage patterns of such equipment over time. When the building’s power measurements are available, machines’ activity state switches over the VLAN can further be correlated to power variations in order to calculate per-machine power consumption.
1.2 Sensor networks
Original purpose
Sensor networks constitute a type of wireless networks with low-range and low-reliability features emanating from the resource-constrained nature of sensor nodes [8]. Applications of such networks range from large-scale environment monitoring to collections of wireless health monitoring devices worn on one’s person [9]. In buildings, sensor nodes are often deployed as stand-alone, e.g. a temperature sensor acting as weather station, but they are increasingly being integrated into building management systems (BMS) [10, 11].
A BMS constitutes in effect a horizontally layered system of sensors, actuators, controllers and user interface devices orchestrated to work together over selected communication media. A BMS may also be divided vertically across differing building subsystems such as HVAC, Fire, Security, Lighting, Shutters and Elevator control systems. Building management systems are deployed in a wide variety of commercial building topologies, including single buildings and multi-building single site environments such as university campuses.
Commercial buildings have been fitted with sensors for over one hundred years. Pneumatics sensors were displaced by simple electronics and dry contacts in the 1960’s. Smart processor based sensors displaced simple contacts in the 1970’s. Recent economic and technical advances in wireless communication allow facilities to increasingly utilise a wireless solution in lieu of a wired solution; thereby reducing installation costs while maintaining highly reliant communication. Wireless solutions are being adapted from their existing wired counterparts in many of the building applications including, but not limited to HVAC, Lighting, and Security systems.
Common sensors found in many rooms include temperature, occupancy, lighting load, solar load and relative humidity. Lighting systems are turned on/off as a function of the overall room light intensity and the presence of persons within the room [11]. Similarly, HVAC systems will also determine the earliest possible time it can shut down heating/cooling yet still control the setpoints to meet the requisite parameters. Besides sensors placed within buildings, weather data reports can be obtained from third party organisations. From high-end weather stations used for professional purpose, e.g. for airport weather monitoring, to any low cost automated weather station purchased by individuals, local conditions can be provided to an online repository. Data readings are generally made available via Application Programming Interfaces (APIs). wunderground.com offers for example an API [12] that developers can use to query local weather data. Files in JSON or XML format are returned and contain reports on temperature, humidity, dew, visibility, wind strength, wind direction, precipitation levels and weather conditions from heavy snow to sunny with more than 15 weather variations. Entries are updated every 30 minutes.
Insights into space and equipment usage patterns
In buildings, sensor networks are generally utilised to optimise occupants’ comfort and experience while reducing electricity expenditure. By purpose they have an inherent link to energy efficiency. For example, controlling lighting and ventilation based on space occupancy does save energy. Other sensors such as temperature sensors may denote whether heating and cooling operates correctly. It is also possible to assess energy losses through temperature readings when electrical heating equipment is powered off. The temperature discharge will indicate how heat is contained within a space. Light sensors can also be used to indicate lighting power state, the latter having a direct link to space occupancy and to electricity wastage.
1.3 Telephone network management systems
In large-scale commercial buildings, telephone networks are generally managed by the IT department. Third party software such as the Cisco CallManager VoIP phone system are nowadays common tools to facilitate management and configuration of telephones. For instance, telephones are generally given extensions that follow a person when that person is moving location from one building to another. Through web management systems, administrators manage names and access levels on phones. Web management of IP phones has the potential to allow central monitoring of space occupancy. Telephone call activity can act as a proxy for identifying whether a given area is occupied, which may in turn be used to produce a space occupancy temporal map to use for electricity wastage identification.
2. ELECTRICAL INSTALLATION
Original purpose
Most large commercial buildings are supplied with three-phase power, but they typically comprise both three-phase and single-phase equipment. Electric meters are instruments generally deployed by electricity providers to measure the amount of electric energy used by a consumer. Meters are calibrated in kilowatt-hours, with one kilowatt-hour the amount of electric energy required to provide 1000 watts of power for a period of one hour. Electricity providers read meters periodically and charge the customer for the amount of electricity used. It is common to sub-meter a circuit to bill usage of a reduced set of equipment. Sub-metering provides an opportunity for building owners and facility managers to identify costly equipment.
The most common type of electricity meter is the electromechanical induction watt-hour meter, which counts the revolutions of an aluminium disc rotating at a speed proportional to the power. New electronic meters transmit readings to remote places and record other parameters of the load and supply such as the load power factor. They are able to support time-of-day billing, where payments relate to the amount of energy used during on-peak and off-peak hours.
Insights on equipment usage patterns and anomalies
Monitoring an electrical installation provides immediate insights on electricity expenditure. Power consumption measurements can be produced for various sub-circuits, direct power measurement is costly and not scalable. More generally, power readings are produced at the distribution board level. Exploitation of electricity data from physical meters refers to the process of generating additional insights from a single power measurement.
Intelligent data processing of electricity readings can for instance enable device-level load disaggregation, but also automated anomaly detection [13] and benchmarking with similar seasons and buildings. More than only extracting additional information from power readings, correlating electrical measurement to concurrent events can provide context and links between power consumption and for instance space occupancy or machine activity patterns.
3. MISCELLANEOUS SYSTEMS
3.1 Machine registration
In the main, machines must be registered first to gain network access in commercial and office buildings. Large organisations such as Universities run networks and have a responsibility to manage such if illegal activity takes place. Discovery and management of viruses require for instance machine inventory and activity tracking. NetReg [14] is often chosen to assist equipment management by providing an online registration process. Through a simple web interface, an unknown DHCP client is prompted for user identification. Powerful scripts retrieve the client’s network finger print and store it along with the user information in a database.
3.2 LDAP access logs
An online directory is a specialised database that stores and retrieves collections of information about resources. Such information can represent any resources that require management, such as employee names and shared network resources such as conference rooms and printers [15]. Online directories can be used by a variety of users and applications, and for a variety of purposes, including looking up e-mail addresses or locating a user’s mail server. LDAP servers are usually deployed for large organisations such as universities and corporations, but smaller LDAP servers exist as well for workgroups [16].
While LDAP can be used to store heterogeneous data, LDAP access logs contain key information valuable in an energy profiling context. Access logs indeed contain detailed information about client connections to the directory [17]. A connection is a sequence of requests from the same client beginning with a timestamp and the client’s IP address. Analysing timestamps and IP addresses of successive connection therefore enables monitoring of machines first and last connections onto enterprise networks. In particular, many corporations require login from network user when network equipment is powered on or unlocked. Such login procedure is generally handled through LDAP authorisation, where a broker needs to determine whether the incoming username is a member of the given group. Utilising LDAP access logs potentially provides reports on employees work hours, but more importantly reports on machine power activity.
3.3 Equipment setback timing strategies
Within buildings that can afford it, building management systems assist facility managers for visualising and controlling equipment operational state from a remote centralised location. Facility managers can set operational schedules via timers, i.e. configure centrally start-up and switch-off times for each piece of equipment, increasing overall control over a site’s energy spending. Optimal operating schedules for HVAC equipment should set start time to be as late as possible and stop times to be as early as possible while maintaining comfort points for occupants. Various degrees of timing control are available, from purely static on and off times, to more complex control algorithms refining operating time schedules dynamically based for example on weather conditions, day of the week, and localised space occupancy.
Depending on the availability of weather and occupancy monitoring systems, timers are continuously adjusted in near real-time or assigned with static values using historical data, models e.g. ASHRAE Standard 90.1 occupancy models [18], and experience from previous installations. Fixed timing strategies are nevertheless often not optimised, since setting timers is a matter unique to each building.
Building managers are in a situation where efficiency of HVAC equipment operation is often unknown and potential savings uncertain, possibly missing opportunities to optimise timer setback schedules. Analysis of HVAC fixed timer strategies and correlation with other data such as space occupancy can potentially be used to uncover electricity wastage.
3.4 Swipe cards
Electronic time clocks automate the collection of employee clockings via swipe cards, and make it easy to monitor employee attendance. Such systems are often used to drastically reduce the manual labor involved in payrolls and control access to restricted areas. Low-end systems do not monitor time and identity of card swipes, but most solutions do produce reports on who entered a given area and at what time. For instance, the swipe card system of the front door of the School of Computer Science & Informatics at University College Dublin keeps track of swipe card activity. In that location, the front door automatically opens during typical working hours (8am—6pm), but requires card swiping outside business hours.
Card holders presence within a building and work hours can be produced down to each individual when digital clock-in is in place. Card swiping provides identity of first people entering and last people leaving a building, and therefore information for portraying a building’s usage pattern and for individualising electricity bills [19].
3.5 Work schedules
Regulations and policies on the administration of work schedules, including the basic number of hours per week and number of holidays, are generally in place within commercial buildings. Categories of employment have however different work obligations. For instance, cleaning staff has to work before a building is occupied, typically from 5am to 8am. Administrative positions have work hours usually ranging between 8am and 5pm every day. Conversely, contract jobs with no regulated work hours do not exhibit such repetitive time pattern. For instance, academic staff and PhD students have varying work hours depending on the workload. Availing of work schedules can therefore provide information about an expected building space usage.
Bank holidays, employees’ individual holidays and meeting room schedules are more accurate resources for predicting space occupancy and expected presence of a given person within a building. Holiday management and room booking systems are sources of space occupancy information that can be used to assess the efficiency of space heating for instance.
3.6 Third party data
Utilising a meter clamped at the power cable is the only accurate method for acquiring the power consumption of a piece of equipment. Sub-metering all equipment within a large-scale organisation is nearly impossible, and other methods must be used to assess loads of individual loads when powered on. Both figures from power specifications/datasheets and third party power measurements on electrical equipment are widely available. Figures from datasheets are usually higher than actual power consumptions. Power specifications are indeed generally worse case power specifications assuming machines operating at full capacity. Third party power measurements may prove to be more accurate for estimating singular power loads since they have been measured with meters during operation. They are however not always available; only the power consumption of common equipment is generally found online.
Utilising third party figures does not ensure high accuracy is estimating power consumption, but using them allows various energy calculations to be produced at low-cost with a certain degree of accuracy.
4. THE CHALLENGE OF INTEGRATING DATA SOURCES
Typical building infrastructure was perused to delineate the energy insights each data source may uncover. The potential of opportunistic data exploitation is real, and the impediments attached to data acquisition from heterogeneous systems must be discussed to position the challenges that need to be tackled.
Data sources will typically range from stand-alone systems with proprietary architectures, e.g. LDAP system, to larger systems encompassing equipment with shared software architecture, e.g. HVAC control system made of compatible sensor nodes and timers, to systems with standard/open architectures facilitating equipment interoperability, e.g. sensor network with ZigBee [20] stack. Major architectural discrepancies between legacy architectures, henceforth closed systems, and systems with standard/open architectures, henceforth open systems, raise integration challenges.
5.1 Open systems
Advantages of architectures facilitating access to raw data are tremendous for electricity profiling based on opportunistic data exploitation. First, access to data at the source allows data from discrepant hardware/software systems to be represented uniformly, through various steps of reformatting. Promoting uniform data representation allows both simplified data interfacing and storage, where discrepancies between each data format would otherwise lead to incompatible system components. Low-level data uniformity furthermore favours data combination and data sharing by supplying a single interface to heterogeneous data. A single dataset can be reused by multiple data analytics software components, and multiple datasets can be used together by a single component. In a closed architecture, a second application would require the deployment of an additional but identical measurement device.
Recent efforts have proposed service-oriented architectures promoting open provision of buildings’ raw data [21], for example a building operating system in which equipment such as sensors and meters provide a service exposing their underlying functionality, e.g. provision of raw sensory and electricity data. These services are accessible through an abstraction layer available either directly at the equipment or through a gateway for legacy devices. Raw data is therefore accessible to a collection of core operating system services and applications. Other efforts such as the European Union FP7 Hobnet project [22] have similarly defined an architecture with data production separated from data utilisation.
Although interoperability and open access to raw data has major advantages, the reality is that leveraging data as a resource of value is often key to establishing and realising strong business cases. Protection against competition therefore reduces the chances of seeing a company sell a product and at the same time allow other commercial solutions to tap in for free into their data sources.
4.2 Closed systems
The great potential of open systems becomes apparent when the system scales in terms of data sources, processing units and applications, and when interoperability is sought. Alternatively, fully integrated solutions provide control over resources. Traditionally, proprietary solutions/products foster vertical architectures, in order that integration of products from other vendors is impossible. For instance, building management systems are often divided vertically across alike, but different building subsystems [10]. Data is seldom openly available and only products from a given vendor can be plugged into each system. Concretely, this means for example that temperature data captured by a Fire sensor cannot be used by a Lighting application. As a result, similar sensors are often deployed in the same location.
Independence from a common architecture provides control over data sources’ resources. A company allowing uncontrolled third-party access to its data sources may indeed endanger the quality of its own solution. For instance, say a wireless sensor network is deployed to report temperature readings every 15 minutes. Allowing other applications to access the network and trigger additional temperature queries may eventually deplete the nodes’ battery lifetime and congest the network.
Data networks, information management systems, building automation systems, and other infrastructure equipment contain unprecedented amount of information. Building automation has not as yet provided the framework that is required for easy information sharing. In the context of opportunistic data exploitation, it is therefore not conceivable that the building infrastructure will be configurable to specific needs. Data will therefore need to be fetched in the format and at the location provided by each data production system.
5. CONCLUSIONS
The infrastructure analysis presented herein confirms that building infrastructures are unexploited sources of energy insights, laying the groundwork for subsequent market research on opportunistic building data acquisition.
The table below summarises energy information that can be exploited from a building infrastructure, reflecting three degrees of energy insights:
-
- Direct insights—produced from continuous analysis of single data streams. Networked equipment power activity states can be inferred at any time from network traffic, whether machines are connected to the network at a given time. The building load demand can be obtained from the building metering infrastructure. The evolution of machines’ connectivity state on the VLAN as well as traces of LDAP connections to shared resources are indicators of electrical equipment usage patterns. Building space occupancy can be inferred from swipe card events and human-initiated activity on the VLAN and LDAP servers.
-
- Meta-information—produced from once-off access to single data sources. Periods of HVAC equipment power activity can be retrieved from BMS time schedules. Data gathered online and from administrative systems will source equipment power characteristics, ownership and location.
- Indirect insights—produced from correlation of multiple data sources. Reports of machines’ periods of connectivity on a VLAN can be translated into load demand and generated costs utilising meta-information. Software load disaggregation applied to both building power measurements and individual equipment power profiles can produce reports on individual equipment periods of power activity and associated load demand. Processing of sensor data combined with building power consumption readings enables measurement of equipment power contribution, matching observed equipment power state with power consumption variations. Space occupancy insights given by individual machines’ connectivity status on a VLAN can be correlated against the building power consumption variations to infer whether BMS timers controlling HVAC equipment power activity are aligned with the building needs.
WHAT’S NEXT?
Wattics is a technology company developing a cloud-based energy management platform, with a mission to assist businesses and energy professionals in turning complex energy data into clear, actionable insights.
The common ingredient and enabler for these applications is the availability and access to data from heterogeneous sources. Wattics works with a variety of data acquisition system manufacturers worldwide, who are experts in specific market domains and applications. This allows the Wattics online platform to seamlessly integrate heterogeneous data streams to uncover insights and serve the market needs.
Get in touch with us to discuss your project requirements so we can advise you the best approach for your energy conservation projects!
+353 1 532 7875 | [email protected]
[1] Electronic Product Environmental Assessment Tool (EPEAT), Available at https://www.epeat.net/
[2] IEEE Standard for Information technology, Telecommunications and Information Exchange Between Systems, Local and Metropolitan Area Network Specific Requirements, Part 15.4: Wireless Medium Access Control (MAC) and Physical Layer
(PHY), Specifications for Low-Rate Wireless Personal Area Networks (LR-WPANs), 2003
[3] G. Kessler, S. Shepard, A Primer On Internet and TCP/IP Tools, IETF RFC 1739
[4] B. Nordman and A. Meier, Using Existing Networks for Energy Purposes, October, 2009
[5] A. Schoofs, A. Sintoni, A. Ruzzelli, G.M.P. O’Hare, NETBEM: Business Equipment Energy Monitoring through Network Auditing, 2nd ACM Workshop On Embedded Sensing Systems For Energy-Efficiency In Buildings (BuildSys), held in conjunction
with ACM SenSys, November 2010
[6] A. Schoofs, A. Ruzzelli, G.M.P. O’Hare, VLAN Auditing for Preliminary Assessment of After Hours Networked Equipment Electricity Wastage, Elsevier Energy, Volume 36, pp. 6910—6921, 2011
[7] R. Melfi, B. Rosenblum, B. Nordman, K. Christensen, Measuring Building Occupancy using Existing Network Infrastructure, Green Computing Conference and Workshops (IGCC), 2011
[8] W. H. Card, A. Sallman, R.W. Graham, E. E. Drucker, Infrasonic Measurement of Building Air Leakage: A Progress Report, American Society for Testing and Materials, Philadelphia, PA, U.S.A., 1980
[9] M. J. O’Grady, P. Angove, W. Magnin, G. M. P. O’Hare, B. O’Flynn, J. Barton, and S. C. O’Mathuna, Enabling Intelligence on a Wireless Sensor Network Platform, 10th International Conference on Practical Applications of Agents and Multi-Agent Systems, Salamanca, Spain, May 2012
[10] J. Martocci, A. Schoofs, P. van der Stok, Commercial Building Applications Requirements, Internet Draft 6LowAPP Working Group, expired on Januar y 8, 2011
[11] D. Delaney, G.M.P. O’Hare, and A.G. Ruzzelli, Evaluation of Energy Efficiency in Lighting Systems using Sensor Networks, 1st ACM Workshop On Embedded Sensing Systems For Energy-Efficiency In Buildings (BuildSys), held in conjunction with ACM SenSys, November 2009
[12] Wunderground.com, Weather API for developers, Available at https://www.wunderground.com/weather/api/
[13] G. Bellala, M. Marwah, M. Arlitt, G. Lyon, C. Bash, Towards an Understanding of Campus-Scale Power Consumption, 3rd ACM Workshop On Embedded Sensing Systems For Energy-Efficiency In Buildings (BuildSys), held in conjunction with ACM SenSys, November 2011
[14] P. Valian, T. K. Watson , NetReg: An Automated DHCP Registration System, SysAdmin: the Journal for UNIX Systems Administrators, Vol. 9 Issue 12, Networking, December 2000
[15] Oracle, Internet Director y Administrator’s Guide, 10g Release 2 (10.1.2), Available at https://docs.oracle.com/cd/B14099_19/idmanage.1012/b14082/intro.htm
[16] IBM Information Center, Configuring authorization with LDAP, Available at https://publib.boulder.ibm.com/infocenter/wmbhelp/v7r0m0/index.jsp?topic=%2Fcom.ibm.etools.mft.doc%2Fap04151_.htm
[17] J. Oakes, E. Deon, D. O’Brien, Red Hat Director y Ser ver 8.0 – Configuration and Command Reference – Edition 8.0.9
[18] American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE), Inc., Energy Standard for Buildings Except Low-Rise Residential Buildings, ANSI/ ASHRAE/IESNA Standard 90.1, 2007
[19] S. Hay and A. Rice. 2009, The case for apportionment, 1st ACM Workshop On Embedded Sensing Systems For Energy-Efficiency In Buildings (BuildSys), held in conjunction with ACM SenSys, November 2009
[20] ZigBee Alliance, Available at https://www.zigbee.org/
[21] S. Dawson-Haggerty, J. Ortiz, X. Jiang, J. Hsu, S. Shankar, D. Culler, Enabling Green Building Applications, Workshop on Hot Topics in Embedded Networked Sensors (HotEmNets), June 2010
[22] HOBNET (Holistic Platform Design for Smart Buildings of the Future Internet) EU FP7 FIRE/STREP ICT Ð 257466, Deliverable D1.3—System Architecture Definition
Blog post image: https://cdn.archinect.net/images/615x/qe/qeptn2zgvsrmkkez.jpg
