using historical data to fine-tune auroraxmp predictive capabilities

13
Using Historical Data to Fine- tune Aurora Predictive Capability and Gain Insights into Market Behavior Chris Handwerk and Derek Salvino October 16, 2009

Upload: epis-inc

Post on 10-Feb-2017

101 views

Category:

Technology


3 download

TRANSCRIPT

Page 1: Using Historical Data To Fine-Tune  AURORAxmp Predictive Capabilities

Using Historical Data to Fine-tune Aurora Predictive

Capability and Gain Insights into Market Behavior

Chris Handwerk and Derek Salvino

October 16, 2009

Page 2: Using Historical Data To Fine-Tune  AURORAxmp Predictive Capabilities

Motivation

Models are great at giving insights, but are only as good as:

Their capabilitiesThe data that you feed into them

AURORAxmp® is an excellent power price forecasting tool, with lots of advanced and detailed functionality.Like a fine racing car, it needs to be tuned up before it can be used to race competitivelyHow do we gain confidence in the model?

1

Page 3: Using Historical Data To Fine-Tune  AURORAxmp Predictive Capabilities

Data 2

http://blogs.smh.com.au/radar/flabbycover.jpg

Having reliable and accurate data is essential Data infrastructure and organization is key

http://celebrity-babies.com/2007/09/09/tina-fey-and-da/

Page 4: Using Historical Data To Fine-Tune  AURORAxmp Predictive Capabilities

Key Input Data (SFELT) 3

Stack (generation characteristics)

Fuels

EmissionsLoad

Transmission

Page 5: Using Historical Data To Fine-Tune  AURORAxmp Predictive Capabilities

Data sources

EPIS databaseExcellent for resource constrained organizations

Publicly availableCEMS (Continuous Emissions Monitoring System - EPA)FERC FormsEIAISO

Proprietary sourcesEnergy VelocityPlatt’sSNLIIR (Industrial Information Resources)

4

Page 6: Using Historical Data To Fine-Tune  AURORAxmp Predictive Capabilities

Suggested Metrics for Comparison – Stack Matching

5

Page 7: Using Historical Data To Fine-Tune  AURORAxmp Predictive Capabilities

Suggested Metrics for Comparison – Unit Operation

6

Page 8: Using Historical Data To Fine-Tune  AURORAxmp Predictive Capabilities

Suggested Metrics for Comparison – Transmission Flows

7

Page 9: Using Historical Data To Fine-Tune  AURORAxmp Predictive Capabilities

Suggested Metrics for Comparison – Zonal LMP (basis)

8

Page 10: Using Historical Data To Fine-Tune  AURORAxmp Predictive Capabilities

Zonal LMP challenges in Aurora zonal

Marginal cost of marginal unit

Bid adderof marginal unit

LMP

Energy

Congestion

Loss

Energy

Between zone loss and congestion

Within zone loss and congestion

Between zone loss and congestion

Within Aurora

9

Page 11: Using Historical Data To Fine-Tune  AURORAxmp Predictive Capabilities

Things worth investigating

Generator characteristicsStart-up costsRamp rates# of bidding segments (upper bidding block)Emissions (seasonality)Outage behaviorMinimum uptimes/downtimesReliability Must Run (RMR)

Transmission link limits/wheeling chargesFuel costs – burner tip v. market price

10

Page 12: Using Historical Data To Fine-Tune  AURORAxmp Predictive Capabilities

Backcasting challenges

How close is good enough?TransmissionFinancial v. physical behaviorBehavior of border regionsIncorporating regulatory considerationsData management and availabilityResources

11

Page 13: Using Historical Data To Fine-Tune  AURORAxmp Predictive Capabilities

12

Finely tuning your racing car will put you in the winner’s circle!

http://www.autobahnmag.com/wp-content/uploads/2009/05/59b48c3fd7circle.jpg