Systems Analysis Advisory Committee SAAC

1
Systems Analysis Advisory Committee (SAAC)

• Monday, March 31, 2003
• Michael Schilmoeller
• John Fazio

2
Agenda

• Approval of the Feb 27 meeting minutes
• The milestones for assessing risk
• A review of the portfolio model
• A review of sources of uncertainties and risk
  mitigation
• Analyzing issues, step by step
• Olivia
• Futures
• Analysis of data

3
Agenda

• Approval of the Feb 27 meeting minutes
• The milestones for assessing risk
• A review of the portfolio model
• A review of sources of uncertainties and risk
  mitigation
• Analyzing issues, step by step
• Olivia
• Futures
• Analysis of data

4
Objectives

• Propose approach to characterizing resources,
  with respect to risk mitigation
• Describe some of the elements we might put in the
  plan arising from our analysis

5
Plan Issues

• Incentives for generation capacity
• Price responsiveness of demand
• Sustained investment in efficiency
• Information for markets
• Fish operations and power
• Transmission and reliability
• Resource diversity
• Role of BPA
• Global change

6
The Issue of Risk

• Variation and uncertainty are critical elements
  in evaluation of each of these
• Assessing reliability is the job of the NPPC, and
  with the exception of wholesale market price risk
  have been significant elements of every plan
• Valuing some aspects risk mitigation was part of
  the third power plan (ISAAC)

7
Sources of Risk

• Fuel price and availability
• Natural gas
• Coal
• Load uncertainty
• Aluminum prices and DSI loads
• Hydrogeneration uncertainty
• Resource availability
• Credit risk
• Extended outages
• Realization of new technologies

8
Sources of Risk

• Transmission congestion
• Exposure to the market and market price risk
• Credit and financial risk
• Emission taxes
• Timing and size
• CO2, mercury, particulates
• Correlation among all of these
• The action of others, such as independent
  suppliers, regulators, etc

9
Temporal Aspect of Risk

• Stochastic, short-term variation
• Long-term uncertainty
• Paths of fundamental prices
• Jumps due to short-term market regimes

10
Risk Mitigation

• Options.
• Right, but not the obligation to take a
  particular action or engage in a particular
  transaction.
• Has two sides, and must be traded between
  participants.
• Usually asymmetric with respect to a given risk
  limits outcome in a single direction.
• Hedging.
• Commitment to action or transaction that reduces
  the variability or uncertainty of outcome. Does
  not provide optionality.
• Usually symmetric with respect to a given risk
  limits outcome in both directions.
• Neither in itself decreases expected costs.

11
Risk Mitigation Optionality

• Long-term flexibility
• Start-up and shut-down speed and flexibility
• Demand reduction
• Mothball and delay flexibility
• Operational and administrative control,
  independence
• Sizing flexibility (capital cost flexibility)
• Short-term flexibility
• Dispatchability, if fixed cost component is small
• Demand curtailment

12
Risk Mitigation Hedging

• Long-term hedges
• Independence from fuel price
• Resource diversity
• High availability and proven technology
• Reliable technology
• Cash flow how and when capital is committed
  (complex)
• RD
• Short-term hedges
• Diversity of fuels
• Reliability of resource and reduced maintenance

13
Risk Assessment Issues

• Technologies and their expected performance and
  economics
• Multiple regions
• Expected variations and seasonality
• Study time period
• Detail and resolution

14
How to Address Risk?

• Quantitative analysis, but in the service of
  insight and communication
• Transparency
• Significance and strategic value
• Value to individual, independent PNW power
  industry participants

Objectives
15
Resource Portfolio
Price-driven generation
Hourly demand
Buy in Market
Sell in Market
Buy in Market
Gas Fired
Hydro
Hydro
Contracts
Total Resources
Coal
Summer
Winter
Summer
Winter
Year 1
Year 2
Background
16

Load Uncertainty Variability
17
How Do We Determine Risk for Our Portfolio?

• Choose a set of correlated values for monthly
  hydro, monthly loads, monthly fuel prices,
  monthly market prices for electricity, etc
• Calculate the cost of our portfolio
• Return to step 1 as many times as necessary to
  obtain a sample distribution that adequately
  describes the underlying distribution of costs
• Apply some metric to the distribution of costs

Background
18
Evaluating the Portfolio
19
Preferred Objective Function

• Preferred risk metric
• Should be a coherent measure
• Subadditivity For all random losses X and Y,
• r(XY) ? r(X)r(Y)
• Monotonicity If X ? Y for each scenario, then
• r(X) ? r(Y)
• Positive Homogeneity For all l ? 0 and random
  loss X
• r(lX) lr(Y)
• Translation Invariance For all random losses X
  and constants a
• r(Xa) r(X) a
• CVAR is preferred candidate

Background
20
Evaluating Cost AND Risk
21
Portfolio Analysis

• Initial questions
• Which measures make sense for the region in the
  short and long-term?
• What kinds of portfolios would benefit, in terms
  of risk management, from which measures in the
  short and long term? How much benefit is there?
• This information provides the rationale for the
  policy choices necessary to achieve
  implementation of preferred portfolios

Background
22
What Is the Tactical Plan?

• Instead of putting all the data into a model,
  turning the crank, and evaluating what comes out
  (good luck),
• Build our regional assessment from the bottom
  up, starting with the most simple elements of
  risk mitigation

23
What Is the Tactical Plan?
24
Example

• Start with wind
• Recognize that we are actually considering any
  resource that is
• Fossil fuel and hydro independent
• Non-dispatchable
• Has a short construction cycle and longer
  shelf-life than fossil-fired generation
• In these respects, resembles solar generation,
  conservation, and so forth

25
Wind Assessment

• Meet 20-year requirement using only wind power to
  a load requirement with expected variation, but
  without uncertainty
• Assume expected reliability for the wind unit
• Assume expected market prices (expected
  variation, but without uncertainty)
• This will represent a benchmark configuration
  about which we perturb our futures

26
Wind Assessment

• Next, vary each source of long-term uncertainty
  in turn. Returning to our list of uncertainties,
  we see these consist of
• Requirement (representing all loads,
  hydrogeneration uncertainty, market exposure,
  unreliability of other units)
• Market price for wholesale power
• Wind resource availability
• Note that we have aggregated sources of
  requirement uncertainty

27
Wind Assessment

• Finally, modify correlations 0.90, 0, -0.90
  among each source of uncertainty, pair-wise.
• So far, this can all be done automatically in a
  workbook model.
• If we expect there are effects that can not be
  captured using pair-wise correlations, we go back
  and re-evaluate those.

28
Wind Assessment

• Now, examine the risk mitigation aspects of wind
  and disaggregate as possible
• Start-up and shut-down
• Mothball
• Sizing
• Note that wind has little or no short-term
  optionality
• Value of hedging is captured via comparing the
  results of this analysis with those of a similar
  analysis of CCCT and other energy sources
  dependent on fossil-fuels

29
Step by Step

• Quantify, summarize, and document
• Go through the same exercise for
• Coal plant
• CCCT
• SCCT/reciprocating diesel/gasoline/propane
• Price responsive demand
• Conservation
• Distributed, local generation vs remote
  generation
• We note that we should be able to evaluate each
  resource independently, as they are constructed
  and operate according to the future they find
  themselves in, not according to the operation of
  any other resource

30
What This Achieves

• We can understand the relative merits of each
  technology by comparing them against a uniform
  benchmark or standard
• For example, we can understand the merits of CCCT
  vs wind generation by examining them
  quantitatively on each dimension
• We can get idea of the desired mix of resources
  by recognizing that each resource effectively
  modifies the remaining system requirement,
  changing its magnitude and correlation with other
  factors

31
What This Achieves

• We are trying to understand the relationship
  between the value of the technology and the
  character of the portfolio in which it functions.
  This exercise addresses that objective directly.
• We are trying to construct an initial regional
  portfolio and discover what combinations of
  technologies is best for the region. This
  should inform our initial estimate and help us
  explain the final outcome.

32
The Prototype Model

• Demonstrated feasibility of running a model that
  runs and optimizes in a reasonable amount of time
• Represents variation in all key variables (CO2
  tax, fuel price, loads, hydro, wholesale power
  market price), large scale uncertainty in loads
  (and fuel price), simplified dispatch, capacity
  expansion, plant reliability

Background
33
Prototype Model
decision variables
correlations and volatilities
conservation assumptions
interest rate, hours per period
chronological structure of uncertainty
conservation calculations
The Portfolio Model
34
Prototype Model
on-peak
off-peak
random variables
input
input
calculation
calculation
The Portfolio Model
35
Prototype Model
annual and study cost calculations and metrics
The Portfolio Model
36
Which Risks Does the Prototype Address?

• We may need greater richness in the description
  of variables, such as separate uncertainty
  forecasts for each fuel
• We may need less richness in other areas, such as
  the subperiods (on- and off-peak) we chose
• New risk mitigation issues that have arise since
  its inception
• Multiple regions and transmission congestion
• Planning flexibility
• DSI load response
• Credit risk/long-term availability
• Availability of new technologies

Background
37
We Could Expandthe Prototype Model

• Pro
• Larger model may be necessary for regional
  assessment
• Con
• Doesnt provide much insight
• Is not a useful to individual participants
• Not as effective for communicating a simple idea

38
Prototype Model

• The prototype model, in itself is transparent and
  a good communications tool, but
• It is fairly rigid, in the sense that it needs to
  be reprogrammed to address special circumstances

Background
39
We Could Write a Separate Model for Each Issue

• Pro
• Can highlight each specific issue and create
  greater insight
• A simple model is a very effective means for
  communicating an idea
• Con
• Very labor intensive
• High potential for errors and inconsistencies
• Keeping track of model runs creates some
  administrative burden, but this is probably not a
  significant drawback
• Still not a useful to individual participants

40
Olivia!

• Olivia writes crystal-ball aware excel workbooks,
  ready for simulation
• Is an integrated database of models
• User can modify a model or create new models
• Of course, the design of all previous models are
  saved, facilitating comparisons and study
  extensions

41
Olivia!

• Pro
• All of the advantages listed above
• Con
• Weird name
• This flexibility comes at the cost of some
  complexity

42
Olivia Demonstration

• Olivia facilitates this process by permitting us
  to quickly and safely zoom our studies up and
  down in detail, without introducing
  inconsistencies
• Olivia.Exe
• Olivia's database
• Olivia.xls

43
Agenda

• Approval of the Feb 27 meeting minutes
• The milestones for assessing risk
• A review of the portfolio model
• A review of sources of uncertainties and risk
  mitigation
• Analyzing issues, step by step
• Olivia
• Futures
• Analysis of data

44
Futures

• System.avi
• Loads uncertainty

45
Futures
46
Agenda

• Approval of the Feb 27 meeting minutes
• The milestones for assessing risk
• A review of the portfolio model
• A review of sources of uncertainties and risk
  mitigation
• Analyzing issues, step by step
• Olivia
• Futures
• Analysis of data

47
Stats Gas and Electric Prices

• Portfolio Model Inputs
• Sumas gas
• Mid-C electric

48
Sumas GasDescriptive Parameters

• Level
• Trend
• Seasonality
• ARMA

49
Sumas GasDescriptive Parameters

• Level, trend, seasonality of ln(Sumas)

50
Sumas GasDescriptive Parameters

• ARMA for residuals

51
Sumas vs. Mid-C

• Seasonality and Trend Removed From Both Series

52
Sumas vs. Mid-C

• Seasonality and Trend Removed From Both Series

53
Mid-C Volatility Seems Periodic

• Variance of Residuals From Mid-C Regression on
  Sumas

54
Next Meeting

• Tuesday, April 29, 930AM, council offices
• Agenda
• More discussion of statistics
• Results with Olivia
• Wednesday, May 28, 930AM, council offices