Smart Buildings as Cyber- Physical Systems (CPS) In Smart Cities: Living Building Dr. Driss Benhaddou Associate Professor and Fulbright Scholar University of Houston, TX dbenhaddou@uh.edu Tuesday, 01/05/2016
Sensor & Wireless Network Driven Intelligent Building Systems
Outline Motivation Living Building Vision and metaphor Description of the system Sensor Information Processing Model Fundamentals Challenges What can we do now Research opportunities
Motivations Smart Grid Energy Efficiency Smart Cities
Our Vision: a Living Building Paradigm A Building/campus/neighborhood that can autonomously manage its power consump,on, genera,on, and control as it is a living organisms. A Building that can op,mize energy consump:on A Building as Microgrid. A Building that is context aware: It interacts with its occupants and adapts to their preferences.
Living Metaphor Nervous System: made of a network (wireless) of sensors, actuators, and compu:ng infrastructure. Regulatory: Building Control algorithms. Immune func,ons: Made of fault tolerant and failure avoidance and detec:on algorithms.
Building as a Cyber-Physical System These intelligent buildings are Cyber-Physical Systems (CPS) that will require a deep integra:on of Ar:ficial Intelligence (Sensing and computa:on), communica:on (e.g. WSN, middleware), and control. These parts are implemented as layered services.
The Physical Layout Smart Buildings With rooftop solar panels Smart Homes Smart Office PHEV Parking Deck MC MC EIS MGCC MC PHEV MC Processor MC MC Renewable Energy Generator Storage Storage Industrial Plant Central Power Plant
Abstraction Layers
Sensed Information Processing Special configurations Repair and maintenance Microgrid maintenance Fault management, etc. Events Production Sensing (solar) Production data from rooftop solar energy Microgrid Management System Adjust temperature setting, machines, etc Adjust microgrid User Feedback Indirect user activity sensing Evaluation Prediction models User activity Models Consumption data from equipment Building Management System Visualization User comfort seting Users activities sensing/ input Consumption Sending
CPS and Smart Grid: Home Scenario
Multi-domain Perspective campus/ neighborhood view End-User Appliance Storage Wireless Link Gateway Internet Campus Net Mesh Node AMI Building Place DataBase Server Application Server Electricity Provider EMS Control Plane Office
SECRETS Sustainable Energy Clusters REalized Through Smart Grids (IRESEN funded) Increase load Charging rate ESS Power House A Power ESS House B Increase load Charging rate Min. Max. Voltages
Middleware Design Management Layer Energy management system Integration Capabilities Business semantics management Security Management Data and system Integration Energy efficiency Demand Response Electric vehicle charging Predictions generation Service Interface Event-based PubSub pattern Configuration accommodation Priorities Assignment Application-Logic Integration Context-awareness Social Networks Service Orchestration: discovery, adaptation... Control Web services Security Logical Interaction (API) Direct Interaction (API) Web Services SOA Web Services development Cloud deployment Sensors/actuators Board Data storage Middleware Data assimilation Data collection Data transforming filtering Data summarization/aggregation Statistics generation device ID Mediation Layer Fault Detection error logs Alarm handling Fault Management Devices AMI T/H sensors, Pressure, Lightning,CO2, Motion. Dampers, AHU, Motors Communication PLC, Modbus,Backnet Ethenet, WIFI, Zigbee, Zwave Abstraction Layer Virtual representation of Network Plug-and-play specification Traffic Regulation Device effecting/actuating TCP/IP
Fundamental Challenges Smart Sensing: smart doors, smart windows, charger, Solar, etc. Network and Complex System Modeling for reliability Human Behavior: Sensing, impac:ng, and Modeling Learning: applied machine learning
What can we do now: Estimation of Building Attendance using WiFi Input: WiFi Network Logs 1200 1000 800 Output: AOendance 600 400 es:ma:on 200 0 Future work: AOendance predic:on based on Markov and/ or regression models. AGendance by Building (11:00am) 409 419 503 505 508 515 516 517 522 547 553 590 1200 1000 800 600 400 200 0 AGendance Building 547
Smart Doors Ultrasound Sensors Infrared Sensors Machine Learning Identification
What Next Dense Sensing Deep Learning Non intrusive Smartness in the network Reliable networks Autonomous building that can act on its own, understands and feels its occupant
Thank You! Question & Answer