A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011...

15
A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1

Transcript of A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011...

Page 1: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

A Complex Event Processing Architecturefor Energy and Operation Management

Jimi WenJuly 14 2011

New York

1

Page 2: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

Outline

IntroductionApplication ObjectivesEnergy Management SystemMaximum Contracted DemandStore Operation

Architecture OverviewRaw EventsPreprocessed EventsRule-based Events

Discussions

2

Page 3: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

Application Context

3

Maximum Contracted Demand ContractKnowledge of these MCD contracts advantageous to utilities. Better estimate on

demand, better planning for generation and transmission infrastructure. Users save on a cheaper plan, if they restrict (Time and Volume) somewhat their

consumption.Total cost = Usage*Rate + Max Usage * Penalty + Adjustment

Store operationMinimization of electricity => decrease costHowever electricity less than 10% of total operation cost, cannot sacrifice operation . But most store main objective is maximize revenue

=> electricity demand is not really elastic. Constraint Optimization (Rules!!)

Page 4: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

Energy Management System

Data CollectionMeters, Sensors, Utilities, Web etc.

Reporting and MonitoringEmission auditing, data visualization, detection, analysis etc.

EngagementConnect users’ energy consumption, advising, feedback etc.

Consumption InteractionScheduling, Bidding, Trading etc.

4

eMeterControl4

AlertMe EnergySeriousEnergy

Page 5: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

Web-Based EMS

5

Deployment6 Stores, 2 Offices, 1 Lab

iMeterData transmitted via ModBus/Zigbee

iGatewayConvert Protocol Stack

Server & DatabaseIntel Xeon 3400 Processor

InterfaceGoogle Map APIData Visualization

Page 6: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

Architecture

Different level languages for different type of users

6

CEP Engine

Rule Engine

ECA Rule

MeterData

WeatherData Event

Handler

EMSSystem

ConsoleOutput

EventLogger

Data A

daptor

Forecasting

Classifier

Windowing

Filter

Plug-in

Page 7: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

Raw Events

Meter DataZigbee communication->gateway->Data serverEach Appliance: voltage, current, power factor, real power, apparent power etc.

Smart Grid DataPrice (For the end-user everything is represent as price?)

Environment Sensor DataMeasurement of site for possible application development?

Web DataForecast weather for solar power?

7

Page 8: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

Preprocessed Events-Forecasted

Preprocessed Events- a kind of plugin written in a more suitable language that can natively express

such computation

Forecasted EventsSimilar time of day approach 15 minutes modelMonte Carlo Method Forecasted probabilities

8

Page 9: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

Preprocessed Events-Classified

Classified EventsDifferent State of Each Appliance monitored

Simple thresholdInference from parametric or stochastic models

Non-intrusive Appliance monitoring

Preprocessed Events Raw, Forecasted and Classified Events are used to together in the rule manager For energy and/or operation management SOP can be easily implemented by non-developers

9

Page 10: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

Rule-based Events

Raw event only: if sensor data dropout, message for system maintenance appointment.

Preprocessed event-classified only: if billboard is not on anytime from 5pm to 5am, alert store manager to switch the

billboard on.

Preprocessed event-forecasted only: If overload probability is high, alert store manager

10

Page 11: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

Rule-based EventsMixture of event types

If overload probability is high and oden3 cooker is on and outdoor temperature is above 30C, then alert to switch o the oden cooker.

"Event:OdenOff" (?Ev, ?T) :-

Exists ?Ev ?O (

And ( "Event:ReadBuffer"( ?Ev "Event:OverLoad" )

"Electricity:ODenCooker"( ?O )

"Electricity:PowerOn"( ?O, ?OOn )

"Environment:Temperature"( ?TM)

"op:greater-than"( ?TM "30" )

"op:text_match"( ?OOn "0n" )

"op:time-now" ( ?T ) ) )

Forall ?Ev ?O (

If And ( "Event:OdenOff"( ?Ev ?T )

"op:time-now" ( ?TN )

"Electricity:PowerOn"( ?O, ?OOn )

"op:text_match"( ?OOn "0n" )

"op:time-after" ( ?TN "5m" ?T)

Then Do( Action(?O "op:power-off") ) )

11

Page 12: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

Discussion

CEP that is specific and embedded How to benchmark?Performance (JSR223+ JSON), not a problem at our scale…Function (JSR 223 + Java)

Enablingness Creating value added solutionsCreate good applications with lesser engine firstBetter interface: Evaluation? Update?

Promising results in 6 store locations

12

Page 13: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

Interface

Which one?

13

Page 14: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

CEP as a platform?

14

Engine- core by big IT provider, IBM, TIBCO etc.Plugins- from niche specialty solutions,Rules- users for rapid development Operation- operation, monitoring, reporting, feedback etc.

Page 15: A Complex Event Processing Architecture for Energy and Operation Management Jimi Wen July 14 2011 New York 1.

Thank you for your attention!!Questions?

15