Navigating the Swamp Scott Eiserloh, Forecasting and Model Development Principal Engineer
Aurora Electric Market Forecasting Conference, Skamania, Sept. 19, 2014
My First Computer
The Early Bird – Analog Economic Dispatch and AGC
Proprietary/Confidential – Not to be shared externally or with SPC
CC
OPCO Plant Type
GPC
APC
GULF
MPC
SPC
H Weiss Dam
Total 50,862 MW
Southern Company “Owned”
N
S
Combined Cycle (12,263 MW)
H
CT
PA
Nuclear (4,715 MW)
Steam (21,382 MW)
Combustion Turbine (5,278 MW)
Hydro (2,595 MW)
Power Purchase Agreements (4,629 MW)
H Lay Dam
H Mitchell Dam
H Martin Dam H Yates Dam
H Thurlow Dam
Sinclair Dam
Logan Martin Dam
H Neely Henry Dam
H Holt Dam
H Jordan Dam
H Bankhead Dam
H Bouldin Dam
H Riverview Dam
H Smith Dam
H Harris Dam
H Rocky Mountain
H Morgan Falls Dam
H Langdale Dam H Bartletts Ferry Dam
H Goat Rock Dam
H Oliver Dam
H North Highlands Dam
H Flint River Dam
Lloyd Shoals Dam
S Plant McManus
Wallace Dam
H Estatoah Dam H Burton Dam
H Terrora Dam
H Yonah Dam
H Nacoochee Dam
H Tugalo Dam
H Tallulah Dam S Plant Gorgas
N Plant Vogtle
S Plant Miller
S Plant Greene County
S Plant Gadsden
S Plant Bowen
S Plant Gaston
S Plant Barry
S Plant Scholz
S Plant Mitchell
S
S Plant Yates
S Plant Scherer
S Plant Hammond
S Plant McIntosh
S Plant Kraft
S Plant Smith
S Plant Watson
S Plant Daniel
S Plant Sweatt
S Plant Crist
N Plant Farley N Plant Hatch
S Plant Branch
CC Washington County
CC
CC CC Plant Theodore
CC
CC Plant Harris CC Plant Franklin
Plant Wansley CC
CC
CT Chevron Cogen
CT
CT
CT Plant Wilson
CT Plant Dahlberg
CT Plant Robins CT
PA Calhoun
PA Mid-GA Cogen PA Wash Co.
(Tiger Creek)
H
PA Solar
H
H
H
PA Monroe/Walton Co.
PA Bainbridge
CT Addison
PA Central Alabama
PA Heard Co.
Plant McDonough CC
CT
CT Cleveland County CT
Rowan County Energy Complex CC
CT
25-MW CCS Demo “World’s largest carbon capture facility on a fossil-fueled power plant”
Alabama Power Plant Barry
Absorber Regenerator
Compressor
Kemper County IGCC Project
Vogtle 3 and 4 Construction
Southern Power Company Southern Power is an unregulated wholesale energy provider. Southern Power and its subsidiaries own and operate 17 facilities in eight states, with more than 8,800 megawatts of generating capacity operating in Alabama, California, Florida, Georgia, Nevada, New Mexico, North Carolina and Texas.
Proprietary/Confidential – Not to be shared externally or with SPC 9
Separation Protocol • FERC ORDER
• Non-Public retail unit specific or transmission
Information Is NOT ALLOWED to be shared with Southern Power
• Public Bulletin Board • SPC Personnel Not Allowed In Retail Work Spaces
– Electronic Security ID Cards • Term Traders cannot know non-public
transmission information • Best Practices May Be Shared
How did we get here?
• Legacy models are many in number and are frequently
being stressed • They include the many in-house developed “tools” • Comfortable, familiar processes are sometimes called
“ruts” • But they usually get us where we want to go
Search for Improvement
• Generalization • Models are developed with a specific problem(s) in mind. • All models are predisposed to look at things in a certain way • This can affect the results • The modeler has to recognize this and sometimes “manipulate or lie” to the model to
get the results that he “knows” are “better”
• What’s out there that may be “better” • Faster • More accurate • More versatile • Easier to use
• “Better” is Defined by the User
• To accomplish a large change, it takes: • Agreement ( buy-in)
• Co-ordinated Effort
• TIME
When Adopting Anything Different, Remember
Aurora Benefits
• Ease of I/O with other analysis tools through Excel • Versatility
– Zonal – Nodal – LT Resource Expansion and Retirement
• Common database and interface – Common data across uses – Facilitates training of personnel – User customizable I/O templates – Scripting capability
• Commitment to Customer Support and Model Improvement is Real
Our Computer Issues
• IT Justifiably Paranoid • Many corporate applications and servers • ~20,000 individual desktops using a common image • Severely restricted admin privileges
• For Aurora We Choose Workstation Over Server • System planning computer farm • “Local” results instead of SQL server • Plan to test solid state drive
Computing Structure
Remote Access
System Planning
Computer Farm
Gigabit Network
Corporate
Desktops for Planning
• Windows 7 Enterprise or Ultimate
• i7-3770 3.40GHz quad core
• Hyper threaded
• 12.0GB Ram
• Dual 22” Monitors
The World According to Our Energy Budget
Current Hourly Modeling • In 2012, over 2,200 Runs • Avoided Cost Studies • Unit Retirement Studies • Renewable Evaluations • Fuel Projections • General Budgeting • Portfolio Analysis • Coal Inventory Management • Gas Hedging • RTP Projections • Financial Planning • Emission Projections • Pool Transaction Projections • Capital Project Evaluations • Outage Optimization
LT Modeling Potential Uses • Retirement Analysis
• Expansion Planning –
Especially Renewables
Nodal Uses • A group within Southern Company is looking at utilizing the Nodal
Analysis capability in Aurora to apply Production Cost Modeling to physical transmission markets. This group is currently evaluating the potential benefits & challenges associated with this analysis approach.
• Potential benefits include: – Identifying system losses – Off-Peak Congestion – Estimates of Forecasted Re-dispatch Due to Congestion
• Useful Aurora Enhancements – Automated Contingency Files – Automatic Constraint Identification Reports – Ability to Model Resource Loss as Contingencies
Term Traders • SWE (Southern Wholesale Energy) is exploring the use of
AuroraXmp Nodal for LMP forecasting and congestion analysis to support various trading strategies in RTOs. The AuroraXmp’s automated scripts have the ability to retrieve market data and update the powerflow model’s load forecast, fuel prices, generator outage periods, and more in several LMP and non-LMP markets. The simulation run times are very prompt and can provide the following output: – Granular Day-Ahead/Real-Time LMP forecast(s) to nodal level – Flow-gate Constraint Impacts to RTO Hub pricing points – Additional Risk Analysis for Hedging Strategies
Southern Power Company and AURORA
22
• AURORAxmp is used by Southern Power’s Market Strategy and Assessment
group to support the development of market price forecasts for the company’s existing and targeted market areas. These forecasts are instrumental in supporting the company’s business development activities and remarketing opportunities.
• Southern Power models the Eastern Interconnect and ERCOT with a custom input database that resides on a SQL Server. The Southern Power base case is a 20 year hourly study.
• Southern Power began evaluating AURORAxmp in 2012 and successfully published its Annual Market Price Forecast using AURORAxmp in July 2014.
Enhancements of Interest • Uplift Logic – making units “whole”
• Convergent EFOR
• Script for Examining ALL Data for a Specific Resource
• Commitment and Dispatch
– Which MIP? – Run Time
• Hydro Scheduling
Convergent EFOR Effects on Dispatch
Input Checking - Single Resource Time Resolution
Call Support for Details
408000
408500
409000
409500
410000
410500
411000
411500
G 0.
001
6 8
G 0.
001
5 2
G 0.
005
6 2
G 0.
005
6 12
G 0.
001
6 1
G 0.
01 7
4G
0.01
5 8
G 0.
001
6 4
G 0.
001
6 3
G 0.
001
3 2
G 0.
001
4 8
G 0.
005
7 3
G 0.
005
7 1
G 0.
01 6
3G
0.00
5 6
1G
0.00
5 4
1G
0.00
1 4
3G
0.00
5 3
4M
0.0
1 3
24G
0.01
6 1
2G
0.01
6 2
M 0
.005
6 8
M 0
.005
5 8
G 0.
005
4 2
M 0
.005
2 2
4G
0.00
5 2
1M
0.0
1 3
4G
0.01
3 8
M 0
.01
3 3
G 0.
005
3 3
G 0.
01 4
1G
0.00
1 2
3M
0.0
05 5
12
G 0.
001
2 2
M 0
.01
4 4
G 0.
001
7 24
G 0.
001
5 24
G 0.
001
6 24
G 0.
001
3 24
G 0.
01 2
3G
0.00
5 4
24G
0.01
2 4
G 0.
001
2 8
M 0
.01
2 4
G 0.
005
7 24
G 0.
01 5
24
G 0.
005
6 24
Production Cost Order
G
M
0
10
20
30
40
50
60
24 8 3 1 12 4 2 24 8 3 1 12 4 2 24 8 3 1 12 4 2 24 8 3 1 12 4 2 24 8 3 1 12 4 2 24 8 3 1 12 4 2 24 8 3 1 12 4 2 24 8 3 1 12 4 2 24 8 3 1 12 4 2
Hour
s for
20
Year
s
Hours Sampled
20 Year Est. Run Times With Gurobi 0.01 2
0.01 3
0.01 4
0.01 5
0.01 6
0.01 7
0.005 2
0.005 3
0.005 4
0.005 5
0.005 6
0.005 7
0.001 2
0.001 3
0.001 4
0.001 5
0.001 6
0.001 7
Data missing due to timing across midnight.
0
100
200
300
400
500
600
120
140
160
180
110
0112
0114
0116
0118
0120
0122
0124
0126
0128
0130
0132
0134
0136
0138
0140
0142
0144
0146
0148
0150
0152
0154
0156
0158
0160
0162
0164
0166
0168
0170
0172
0174
0176
0178
0180
0182
0184
0186
01
Run-Of-River Hydro
Output
Capacity
Set Energy Shift Method = 0 To Get Fixed Output
0
5000
10000
15000
20000
25000
30000
35000
40000
45000
0
200
400
600
800
1000
1200
1400
1600
18001 19 37 55 73 91 109
127
145
163
181
199
217
235
253
271
289
307
325
343
361
379
397
415
433
451
469
487
505
523
541
559
577
595
613
631
649
667
685
703
721
739
Weekly
Monthly
Demand
Peak Shaving Hydro Example
Retirement Analysis • Appears to meet our needs, but has not been sufficiently explored
– Dilemma – what is fixed and what is variable? – Must appropriately structure the data to fit Aurora – Some important items affecting resource retirement decisions
are not directly quantifiable • Personnel • Local economy impacts
– Method and results must meet scrutiny of PSC.
Current Approach for LT Expansion Planning
Area 1
Area 2
Unlimited Flows
Retail Only Expansion • DSO’s – Resource; 100% @ Peak, But Always on Maintenance • Purchase Transactions Modeled as Resources • Some resources which are obligated to third parties are not
accounted for in the expansion plan. • A plan may be “profitable”, but can it also be least cost to retail • Stop on least cost option is not the same as least production cost
LT Issues Continued • Can over-build relative to annual reserve margin requirements be
least cost? – Might this result in a lower total production cost? – Even if it does, there could be near term rate impacts. – New option to “minimize reserve margin” may address this issue. – Can we get comfortable with the different way end effects are handled – Are the solutions “close enough” compared to what we are used to?
CONCLUSION
• EPIS Commitment to Customer Support and Model Improvement is Real
QUESTIONS?
Proprietary/Confidential – Not to be shared externally or with SPC 37
Nacogdoches Generating Facility – 100 MW
Proprietary/Confidential – Not to be shared externally or with SPC 38
CIMARRON SOLAR FACILITY – 30 MW
Lloyd Shoals Dam – 14 MW Capacity – 6 Units Martin Dam – 186 MW Capacity – 4 Units
Wallace Dam 321.3 MW – 6 Units, 4 Reversible for PSH Rock Mountain PSH Facility – 1095 MW with 613 Foot Head