Modeling Regional Electricity Generation

1
Modeling Regional Electricity Generation

• Phillip Tseng
• Statistics and methods Group
• Energy Information Administration
• April 2007

This is a working document prepared by the
Energy Information Administration (EIA) in order
to solicit advice and comments on statistical
matters from the American Statistical Association
Committee on Energy Statistics. This topic will
be discussed at EIA's spring 2007 meeting with
the Committee to be held April 19 and 20, 2007
2
Overview

• Review of past effort
• Objectives of this modeling effort
• Data Sources
• Challenges
• Methodology
• Preliminary findings
• Questions on fuel switching for the Committee

3
Past Effort

• Started regional electricity modeling about 2
  years ago.
• The objective was to increase regional details to
  the short term projections and analysis
• The methodology used variable costs to represent
  supply curves, required longer solution time, and
  was very difficult to achieve satisfactory
  solution
• Lost a valuable colleague

4
Objectives

• Develop a short-term regional electricity model
  capable of performing two functions
• Project power generation from power plants
  burning fossil fuels
• Conduct scenario analysis and provide insights
  into potential power flow problems
• We focus on fossil plants because in the short
  run dispatching pattern of other types of power
  plants are less responsive to changes in demand.

5
Data Sourcesby state/region
Net generation by generator/plant EIA-906 Monthly
(calendar)
Sales and revenue by end-use sector EIA-826 Month
ly (billing)
Power control area Daily Hourly Load
Generation capacity by fuel type EIA-860 Annual
Demand curves (24-hour load)
Supply curves (By fuel type)
6
Electricity Regions/Sub-regions

• New England
• Mid Atlantic less New York
• East North Central
• West North Central
• South Atlantic less Florida
• East South Central
• West South Central less Texas

• Mountain
• Pacific less
• 10. California
• 11. Florida
• 12. New York
• 13. Texas

7
RSTEM 13 Electricity Demand and Supply Regions
8
Challenges

• Develop a methodology to integrate these data
  with different characteristics
• Provide a reasonable modeling framework to
  project generations of fossil plants
• Provide an analytical framework capable of
  answering what if questions
• Use Eviews to solve and maintain the model with
  ease

9
Dispatching Methodology - 1

• Use most up-to-date data (year 2005) to project
  generation for the next 24 months
• Power plants owner should dispatch least cost
  generators first then move to the next lowest
  cost units as demand increases
• Actual dispatching decisions may depend on the
  least system cost instead of least variable cost
  because of spatial considerations such as
• costs and accessibility of transmission lines,
• distance to load centers,
• transmission and distribution losses,
• and availability of fuels.

10
Dispatching Methodology - 2

• The construction of short-run supply curves for
  coal, natural gas, residual fuel, and diesel
  fuel, uses observed dispatching patterns instead
  of costs due to lack of detailed data.
• Fossil plants are sorted by utilization rates and
  assigned 10 bin numbers
• Two functional forms are selected to fit the
  supply data (cumulative capacity) one for each
  of the 4 types of fossil plants (coal, gas,
  diesel , and residual fuel)
• For each demand level, adjusted for non-fossil
  plants, generation is determined by equating
  demand to the combined quadratic supply curve.

11
Bin Assignment by Fuelsfor Texas (capacity in
megawatt)

• Capacity by bin numbers Cumulative
  capacity
• Bin Coal Diesel Resid NG Coal Diesel Resid NG
• 1 4623 1 0 2319 4623 1 0 2319
• 2 3944 2 0 5667 8567 3 0 7986
• 3 6670 3 0 6647 15237 6 0 14633
• 4 2764 4 0 6945 18001 10 0 21577
• 5 0 6 0 5313 18001 16 0 26890
• 6 0 9 0 5333 18001 25 0 32223
• 7 1880 12 0 3052 19881 37 0 35275
• 8 0 14 0 8714 19881 61 0 43989
• 9 0 15 0 4816 19881 66 0 48805
• 10 0 16 0 14461 19881 82 0 63265

12
Adjust Availability Factors of Coal Plants

• Most operators of coal plants reported use of
  small amounts of gas, diesel, or residual fuel.
  It implies that coal plants are regularly shut
  down for ash removal or routine maintenance. For
  model calibration, we applied three factors to
  three regions
• Eastern region 0.93
• Western region 0.94
• Texas 0.96

13
Fit the Curves

• For coal the functional form is
• Log(coal) C(0) C(1) (1/bin)
• For Gas, diesel, and residual fuel, the
  functional form is
• Log(XXX) C(0) C(1) bin C(2) Bin2
• Natural log is used to avoid negative values.

14
Texas Coal Supply Curve from July 2005
Data(right axis is log of capacity, horizontal
axis is bin )
15
Texas Gas Supply Curve from July 2005 Data
16
Texas Diesel Fuel Supply Curve estimated with
adjusted July 2005 Data
17
Convert Monthly Sales to Load Curves
O
18
Subtract non-fossil generation from load curves

• For nuclear, renewable energy, and other
  non-utility fossil generation, we divide the
  generations by 24 and subtract them from the
  hourly loads. For solar and wind, we can change
  the adjustment to the load curve easily.
• For California, we assume hydro generation is
  proportional to the 24-hour load curve
• For hydro storage, we divide the number by 6 and
  subtract from load 1 through 6. (note hydro
  storage is reported as negative)

19
Solve for Generation
Total fossil supply
gas
coal
oil
Bin
Generation
demand
20
Preliminary Findings

• The 2005 monthly dispatching curves worked well
  for coal and gas generators in most months
• The out of sample test perform satisfactorily for
  historical 2004 and preliminary 2006 data
• The model is capable of capturing seasonal
  fluctuations
• Shape of load curves (max and min, kurtosis) may
  affect dispatching of oil-based generators.

21
Eastern Region Coal and Gas Generation(mwh/day)
22
Western Region Coal and Gas Generation(mwh/day)
23
Texas Coal and Gas Generation(mwh/day)
24
Performance of the Model - 1

• We use two measures to evaluate the performance
  of the model
• Mean absolute percentage error
• Twelve month percentage difference

25
Performance of the Model 2(mean absolute
percentage error)
26
Performance of the Model 3(12 month percentage
difference)
27
Modeling Issues

• The electricity dispatching model does not do a
  good job in capturing dispatching of oil-based
  generators.
• The dispatching curves assume system costs do not
  change from one year to the next. It needs an
  adjustment factor to consider oil and gas price
  changes
• More generation will be from gas units when gas
  is cheaper
• For dual fueled generators, some units may switch
  from gas to oil when gas is cheaper

28
Questions for the Committee

• A few sub-regions in the East have generators
  that can switch from gas to oil or vice versa.
  The adopted modeling methodology needs to
  estimate a switching variable to capture the
  effects of gas and oil prices on gas and oil use
• Is it appropriate to use sub-regional generation
  data and fuel prices to estimate the effects of
  prices on fuel switching?

29
Simulation and Contingency Analysis

• Simulate the effects of power outage on
  generation and inter-regional power flow
• Simulate the effects of gas supply shortages on
  regional generation
• Simulate the effects of emissions constraints on
  generation (we can truncate the supply curve)
• These simulations can be performed by changing
  the capacity matrix and re-estimating the supply
  curves

30
Sub-regional Demand and total generation in the
Western Region (meg watts per day)
31
Insights on Transmission Problems

• A well-calibrated model can show load,
  generation, and imports/exports in each
  sub-region
• The simulated hourly indigenous generation and
  imports can provide valuable information on
  supply capability and possible bottlenecks on
  imports
• We use California as an example to demonstrate a
  potential application of the model in this area.

32
(No Transcript)
33
(No Transcript)
34
Simulation and Contingency Analysis

• Note that every capacity matrix is derived from
  EIA-860 and EIA-906 and can be traced back via
  plant ID.
• Each capacity matrix has a corresponding fuel use
  matrix and emissions matrix, which can be used
  for contingency and emissions analysis.
• These types of analysis can be performed
  iteratively by running the model, assessing the
  fuel use and emissions, and adjusting the
  capacity matrix and power supply curves.

35

• Thanks for your Attention