TOWARDS LIVE, CONTINUOUS BUILDING ENERGY AUDITS Recording and Managing Building Plug-load Information Jorge Ortiz and Jason Trager LoCal Retreat June 1, 2011
Many miscellaneous loads Plugs loads can account for almost the other 50% of energy spend in the building Opportunity for savings: Plug loads may be controlled locally either by user of by demand-response signal Metering and collecting this data takes long Many sources distributed throughout the building
Energy audits today Can range from simple building walk-through to analysis of energy use (electric bill) Comparison made to similar buildings Most sophisticated audits may use install metering and offline building modeling
Rapid Auditing Protocol (RAP) 4 Tag everything with a QR code Scan Device QR code (unique for each device) Bind QR code with device/item Bind meters and devices Record Device information Performed audit in Sutardja Dai Hall
Recording spatial context Global Origin (3,0,0) Building coordinate system (foot by foot) Origin: Southwest corner of SDH
RAP Interface 6 Android App Easily used Shallow learning curve (356MN,8,2) (356MN,8,8)
Environment and Activity /SodaHall /hvac /CT /Chiller /loadtree /panel /xform /spaces /floor3 /floor4 Electrical Load Tree Climate plant StreamFS: Metadata organization 7 Structured, Traversable namespace Multiple building views
StreamFS: Query system architecture 8 /inventory/mote123 DB PID 1 PID 2 PID 3 PID 4 Time GET ?query=true&ts_timestamp=gt:now-100,ls=now GET /SDH/spaces/*?query=true&props_metertype=powermeter { “metertype”:”powermeter”, “desc”:”Electric power meter”, “timestamp”: } /power /temp/hum/par
Audit metadata organization /is4 /buildings /taxonomies /SDH /qrc /mels /Electronics /Miscellaneous /other /inventory /spaces LBNL Taxonomy
Recorded Information Basic info Power Usage, placement, device power draw or amperage rating, device make and model number Device Type (taxonomy) Provided by LBNL Extra information
Initial Rapid Auditing Results TypeComputerElectronic Desks LampLaser Printer LCD Monitor Notebook Computers PhonesImacResistiv e Loads MicrowavesOther Count Power Use
RAP / StreamFS Data pulled from StreamFS Building Rooms Plotted Static Power use Easily understood Tap into people’s perceptions Give estimations of energy use
Upcoming work Map to other building views (electrical, HVAC) Track deployment evolution Integrate live metering Tap into the “Weak Actuator” – People Live alerts and visualizations Address Coverage How many sensors are needed to capture energy picture? Use audit to investigate DR Variations How is our scenario unique?
Tracking deployment evolution Metadata organization Typically static, now live. Streaming data from meters Items attached to those meters changes Meters move as well Moved to new locations Attached to different items How do we keep accurate view of where energy is going and how it is being consumed? Context-related queries: One-off queries on current state Live queries that account for state transitions
Metadata graph Acyclic graph of namespaces for building views /SDH /inventory { “desc”:”invento ry inside SDH” “timestamp”: … } {“desc”:”Lamp” “timestamp”: …} {“desc”:”Phone ” “timestamp”: …} {“desc”:”Outlet ” “timestamp”: …} {“desc”:”Acme” “timestamp”: …} r-node s-node /hvac /CT /Chiller /vent /loadtree /panel /xform /outlet /spaces /floor3 /floor4 /power /mote123
Typical subscription /buildings/SDH/floor3/364/LCD/power | String match on source data Data POSTed to ‘ /buildings/SDH/floor3/364/LCD/power ’ tagged on entry Tag used to match against subscriptions and pushed to matches
Dynamic subscription /buildings/SDH/floor3/364/* | Still a simple match of the wildcard source string What about relevant data that whose ingress tag does not match? i.e. If a mobile power meter is measuring LCD in room 364 and it is placed in another folder (i.e. /inventory) Incoming data tag will not match wildcard expression POST /inventory/meter
Dynamic subscription (2) Using the metadata graph Check match and path existence /buildings/SDH/floor3/364/LCD/power (symlink) /SDH /inventory { “desc”:”inve ntory inside SDH” “timestamp” : … } {“desc”:”La mp” “timestamp” : …} {“desc”:”Ph one” “timestamp” : …} {“desc”:”Ou tlet” “timestamp” : …} {“desc”:”Ac me” “timestamp” : …} r-node s-node /hvac /CT /Chiller /vent /loadtree /panel /xform /outlet /spaces /floor3 /floor4 /power /mote123 /LCD
Traditional RDBMS vs StreamFS Dealing with static data …but stuff moves around Must re-run query as changes happen Distillation scope change over time The components accounted for changes with context Aggregate data changes in accounting w/evolution of deployment Not just on/off, but there or not there Push vs Pull Data pushed to external targets User may install static or dynamic subscriptions to incoming data
Related work Traditional energy audits Auditing MELs Recent LBNL study Items recorded by hand Live metering with ACme power meters integrated with the deployoment sMAP + typical RDBMS (MySQL) and non RDBMS (MongoDB) Data collection framework Taxonomy integration
Related work (2) Streaming queries and subscriptions Pub-sub information bus Data items self describing and include topic for consumers to subscribe to StreamFS uses pathname and item properties as topics, but also account for inter-relationship through metadata graph Unix FS + pipes Typical hierarchical namespace with symbolic linking Pipe abstraction for subscription installation
Thanks Questions?
Extra