1
Complex Event Processing for Intelligent Energy Management in Microgrids http://smartgrid.usc.edu/ Zach Gima, Qunzhi Zhou, Yogesh Simmhan and Viktor Prasanna Target Application Introduction Past & Present Experiments Pattern Detection Results Present • Classroom Scheduling vs. Building Energy Load • Load Curtailment Potential for Buildings with many Air Handling Units (AHU) Past • Model meaningful dynamic DR situations using complex event queries • Validate real-time DR pattern detections Discussion & Future Work Acknowledgement This work is supported by the Energy Informatics Center at the University of Southern California, Facility Management Services (FMS), Los Angeles, and its funding agencies, the US Department of Energy (DOE) and Los Angeles Department of Water and Power (LADWP). Special thanks to Carol Fern and Roddy Lee from FMS, Nicole Sintov and Mike Orosz from the ISI, and students in the Smart Grid group at USC. • CEP can account for inconsistencies in some of our prediction models • We will be exploring different CEP taxonomies to assist in dynamic DR optimization in microgrids. • Coupled with DR, CEP could allow for Automated Demand Response, lessening the responsibility of system operators • We plan to look into using CEP to send real time notifications to energy managers as well as use the smartgrid.usc.edu portal to display complex events in real-time Solution Approach Technology – Complex Event Processing • Primitive event – information at a point of time, e.g., meter and sensor readings, class schedules • Complex event – patterns of primitive events in a time window, e.g., sequence of meter readings • Complex event pattern detection Traditional DR • Schedule based • Incentive based e(t) time t pattern {e(ti) | t1<ti<t2} time t1 t2 complex events primitive events Pa#ern Queries sta-c meta data D 2 R Semantic Complex Events (Event Patterns) Pattern 1: The 5-minutes average consumption of MHP exceeds its pre-peak demand 27KW. Pattern 2: A Meeting Room’s temperature is above 73 degree when it is not occupied. Pattern 3: More than 6 Fan Coils in MHP operate at the same time. Pattern 4: EE Department’s power consumption exceeds its pre-peak demand 600KW. data streams Smart meters HVAC sensors Appliances D 2 R Pa#erns Complex Event Processing Engine System Overview • Complex event processing engine – correlate both static and dynamic information • Static meta data – domain knowledge such as sensor locations, room types and so on • Data streams – real-time data from meters, HVAC units and appliances on campus • D 2 R Patterns – CEP queries categorized as monitoring, prediction, and curtailment patterns LADWP’s Renewable Portfolio Standard: Challenges and Implementation, Robert Hodel, DWP (CEP) Campus Microgrids Heating, Ventilation, and Air Conditioning (HVAC) Smart Meters Fan Coil Unit (FCU) Diversity •Large Volume and fine-grained data •Varied information sources •Multi-disciplinary participants Evolution •Evolving infrastructure •Changing consumer profiles and preferences Energy Management Challenges Dynamic DR (D 2 R) • Adaptable • Opportunistic curtailment • Low latency Demand Response Optimization (DR) • Extensible application modeling • Real-time data processing • Heterogeneous information correlation • Diverse and evolving client USC Smart Grid Portal Energy Heatmap 0 5 10 15 20 25 30 35 40 12:00:00 AM 1:00:00 AM 2:00:00 AM 3:00:00 AM 4:00:00 AM 5:00:00 AM 6:00:00 AM 7:00:00 AM 8:00:00 AM 9:00:00 AM 10:00:00 AM 11:00:00 AM 12:00:00 PM 1:00:00 PM 2:00:00 PM 3:00:00 PM 4:00:00 PM 5:00:00 PM 6:00:00 PM 7:00:00 PM 8:00:00 PM 9:00:00 PM 10:00:00 PM 11:00:00 PM MHP Duty Cycle Data Previous Curtailment Data
