Automating Campus Dashboards Using Haystack
TRACK 1: Moving to High Performance Data Driven Buildings
Stephen M. Frank, PhDNational Renewable Energy Laboratory
May 13-15, 2019 2
1. Understand how Haystackenables dashboard automation
2. See some practical examples ofHaystack-enabled dashboards
3. Review challenges and solutionsfor dashboard automation
Learning Objectives
May 13-15, 2019 3
NREL Intelligent Campus TeamSuzy Belmont
Information Systems Coordinator
Kiley TaylorMetering Infrastructure
Engineer
Dylan CutlerOptimization
Engineer
Stephen FrankInstrumentation & Controls Engineer
Lissa MyersAnalyst
William GilliesUX/Brand Designer
David GoldwasserBuilding Energy
Modeler
Tau KungBuilding Energy
Research Engineer
Sakshi MishraMachine Learning
Specialist
Alex SwindlerSoftware Engineer
Bethany SparnSystems Engineer
Michelle SlovenskyProgram Manager
Jacob ReynoldsonControls Technician
Anya PetersenSoftware Architect
May 13-15, 2019 4
Circa 2003Facility Floor Area: 323,776 ft2
Occupants: ~450
Circa 2018Facility Floor Area: 967,390 ft2 (+198%)Occupants: ~2100 (+367%)
5
Electricity Meters
NREL’s Energy Management & Information System
Building Automation
Systems
Onsite Generation
Electric Vehicle Service Equipment
Weather Measurements,
Forecasts
23 Facilities
Complete In Progress Meters Only No Data
22,842 Data Streams
5 2 13 5
May 13-15, 2019 6
NREL EMIS Architecture
BAS
NREL-OwnedMeters
Utility-Owned Meters
SkySpark
Fault Detection
Dashboards
EVSE Demand Management
Other Research
Reporting
Weather Station
Electric Vehicle Service
EquipmentVarious Clouds
BACnet Connector
Modbus Connector
Custom Connector
CustomIntake Scripts
HistorianDatabase
Metadata Database
Research Servers
Web Services
EVSE Control Interface
DevicesApplications
Single API
API Load Balancer
Custom API
BACnet IP
Modbus IP
Proprietary
Web APIs
NREL Research Control Protocol
Haystack API
SkySpark Internal
Proprietary
Provide informationto campus occupants
May 13-15, 2019 7
Provide visibilityto facility operators
Provideinsight for businessdecision makers
NREL’s Dashboard Goals
May 13-15, 2019 8
Most dashboards still use manually-configured communication links
Image: Joseph A. Carr, Wikipedia (used with permission)
May 13-15, 2019 9
Our design philosophy:Automate dashboard generation using Haystack
When an operator modifies facility information…
…all dashboardsupdate automatically
DynamicNavigation
May 13-15, 2019 10
Goal: Dynamic
Everything!/nav
DynamicNavigation
DynamicLayout
DynamicData
Benefits:
Portability
Scalability
MinimalMaintenance
Q. How does it work?A. Tie modular UI elements
to Haystack queries
May 13-15, 2019 11
Query: site
{site, dis:"Cafe", …}
Autogenerated Navigation:http://intelligentcampus.nrel.gov/#/building/cafe/
Same data, excluding siteMeter, but different view
{site, dis:"Cafe", area:12140ft², …}
1. Query: elec and meter and siteRef==@cafe
3. Read History
2. For meters, query:active and total and power and point
1. Query: point and weatherPoint and weatherRef==@cafe->weatherRef and (temp or ghi or cloudage or …)
2. Read: curValon each point
Number and placement of gauges from query result (using pre-defined layouts)
Expectation models built automatically from historical meter data
May 13-15, 2019 24
Prediction Points
navName: “Real Power Total”id: @p:nrel:r:4point: sensor: active: power: total: unit: “kW”equipRef: @p:nrel:r:3 “Main Meter”spaceRef: @p:nrel:r:2 “Meters”siteRef: @p:nrel:r:1 “Cafe”…
navName: “Real Power Prediction”id: @p:nrel:r:5point: prediction: predictionOf: @p:nrel:r:4predictionAlgorithm: “ANN"predictionConfig: { … }equipRef: @p:nrel:r:3 “Main Meter”…
navName: “Real Power Upper Bound”id: @p:nrel:r:6point: prediction: predictionOf: @p:nrel:r:4predictionAlgorithm: “ANN"predictionConfig: { … }ub: // Upper BoundequipRef: @p:nrel:r:3 “Main Meter”…
navName: “Real Power Lower Bound”id: @p:nrel:r:7point: prediction: predictionOf: @p:nrel:r:4predictionAlgorithm: “ANN"predictionConfig: { … }lb: // Lower BoundequipRef: @p:nrel:r:3 “Main Meter”…
Challenges and Solutions
May 13-15, 2019 25
May 13-15, 2019 26
Challenge: Integrating Spatial Data
• Haystack has limited native support forgeospatial or 2D/3D asset information
• (Partial) solution: cross-reference to anexternal graphical asset store
• Drawback: still requiresupdates in two places
May 13-15, 2019 27
Challenge: API Limitations
• Client must initiate communication:Many clients ⇒ many API calls
• hisRead op is one point only:Many points ⇒ many API calls
• Limited/unclear support for COV subscription:Polling required to get updates (even with watchSub?)
May 13-15, 2019 28
Challenge: Manage Caching
May 13-15, 2019 29
(One Possible) Solution: Middleware
Haystack WebSocketsDatabase
Middleware
Clients
May 13-15, 2019 30
1. Manages rescan/refresh2. Caches asset lists, metadata, navigation3. Centralizes updates ⇒ reduces API traffic4. Pushes new data to clients
Roles of the Middleware
May 13-15, 2019 31
Challenge: Filtering / Visibility Control
Sometimes, you don’t want everything to show up…• During setup/commissioning• Important sites only• Sensitive facilities
(One Possible) Solution: Filter with dashboard tag
Live DemoTime and Technology Permitting…
May 13-15, 2019 32