|
Summary |
Given the move toward competitive power markets, this project is examining the incentives under market-based decision-making for market participants to invest in the careful design, operation and maintenance of the overall power system. Market designs will be proposed for ensuring power system enhancement under a wide range of operating condition and market forecast uncertainties. |
|
Research Need |
As the US moves towards competitive markets in electric power generation, the decoupling of responsibility for maintaining system security from the ownership of facilities provides a source of new problems. The natural question arises as to what contract terms and incentives will exist for all agents to invest engineering time into the careful design, operation, and maintenance of the overall system. This question becomes more pressing if the added operational uncertainties in a restructured environment make generation scheduling, demand scheduling, and operation more difficult. Note that most decisions have significant impact on the scheduling of the system, and therefore significant impact on the economic "bottom line." |
|
Research Stem |
Markets |
|
Academic Team Members |
Gerry B. Sheblé (Iowa State University – lead: gsheble@iastate.edu), Daniel Berleant (ISU) and Robert J. Thomas (Cornell) |
|
Industry Team
Members |
Dale Stevens (MidAmerican) |
|
Funding Period |
September 2000 to March 2003 |
|
Budget |
$70,000 per year for 2000 and 2001 carried forward |
Project Description:
As the US moves towards competitive markets in electric power generation, considerable effort has been focused on issues of ancillary service markets, and potential system security consequences of differences between contracted services and the physical supply of power services. Such service requirements introduce uncertainties that must be included in the security of delivery. These issues relate primarily to the contingent state behavior of the system, on time scales of minutes to hours. The impact of changes and uncertainties in system operation on contingent contracts has also, to date, been largely overlooked.
This project is intended to achieve two goals. First will be the characterization of market interaction between primary energy exchange and reserve margin requirements as driven by market actions given market forecast uncertainty. Second will be proposals of market power philosophies, and specific market designs, that offer quantitative schemes for ensuring system enhancement under a wide range of operating condition uncertainties. The theoretical underpinnings of such an approach are known in the economic and market literature. It is primarily proposed to evaluate the effect of uncertainties in the price bidding and matching of such systems for such an array of contracts. Some the recent tools of probability and statistics theory, such as interval analysis, and parametric programming will be brought to bear on these problems.
Potential Industry Benefits: Standardized market interaction and base line strategies to facilitate secure addition of competitive resources from all agents units, including widespread distribution level equipment. The benefits of properly assessing uncertainties, even when such uncertainties can be measured, enable the proper pricing of the primary product, energy, and all supportive, ancillary services.
Expected Outcomes: Biannual project reports on progress. Prototype designs for market models for primary energy delivery and for contingent services (spinning reserves, regulation, load balance, ready reserves, and credit reserve) for competitive agents. Test-bed system indicating system benefits of these uncertainties evaluation tools when interacting with competitively driven bidding of energy dispatch and ancillary service provisions.
Technical Approach:
The objective in previous studies was to guarantee that market risk was not dominated by any single contract. Clearly, contract behavior and control challenges arise as large numbers of independent, for profit companies trade for risk management based on the latest information of the equipment and the capability of system operation. The project will consolidate and extend these preliminary results, considering their application to more widespread supportive markets such as spinning reserves, ready reserves, load balance and regulation markets. This project will tailor auction and bidding designs to a competitive power system environment, as well as examine results in trajectory sensitive markets as contracts change based on new information.
The
technique we will use, IBDO, discretizes input distributions by using a list of
numerical intervals that span the range of the distribution and associating
with each interval the probability that a sample drawn from the distribution
would fall within it. A relatively straightforward convolution process does a
sum, product, or other operation on two inputs that are known to be
independent. In the case of unknown dependency, an optimization process is used
to calculate strategic points on the envelopes describing the outputs. This
process (Berleant and Goodman-Strauss 1998) is both simpler and more flexible
than a technique addressing the problem described by Williamson and Downs
(1990).
Work Plan:
|
Task |
Description and Completion Date |
|
1 |
6 months: Extend ISU contract models with existing contingent strategies for risk management. |
|
2 |
9 months: Augment model system to examine significant uncertainties at transmission and distribution levels as well as generation and demand levels. |
|
3 |
12 months: Examine interaction of generator company and energy service company contracts with transaction-based unit commitment and dispatch algorithms. First proposal for active bidding designs for generation companies in competitive generation units. |
|
4 |
18 months: Prototype "cooperative" contract strategies for generation companies, integrating ISO operational decisions with local decisions of unit owner. |
|
5 |
24 months: Prototype "cooperative" contract strategies for energy service companies: integrating enhanced configuration estimation for market analysis, incorporate ancillary services of demand responsive to supportive services. |
Related Work:
Various PSERC teams have and are investigating the impact of market designs and of market interaction with the network. Such work includes the following:
§ Assessment of Transmission Constraint Costs: Northeast U.S. Case Study
Such projects are related to this project but do not
include uncertainty on a contract or market basis while considering multiple
market efficiency. Risk management requires efficient contract trading through
efficient, liquid markets.
How This Work Differs
From Related Work:
Many
useful calculations are simple when we know precisely what the values of the
variables are. However, when we apply such calculations to real world problems,
precise values are frequently not justified by the knowledge available, and
indeed may not be practically obtainable at all. Too often, the approach taken
in such cases is to use precise values despite there being only approximations
of unknown actual values. While this allows easy calculation of numerical
outputs, those outputs generally deviate to a greater or lesser degree from the
true values. When the implications of an inaccurate output are significant, as
when significant amounts of financial or other resources are involved, such
approximations can be costly; on the other hand, techniques that account for
uncertainty in the inputs can lead to significant savings.
Diverse
techniques have been described for calculating with variables that are
uncertain. Ideally, a technique should avoid approximating distributions used
to describe uncertain inputs, because avoidance of approximation is the reason
for applying such a technique in the first place. A technique should also be
robustly applicable, and not require that uncertainty be expressed using
Gaussian or other predefined distribution types. Finally, a technique should
avoid assuming that input distributions are independent of one another or have
any other predefined dependency relationship, a non-trivial requirement. These
factors lead us to propose using the technique of interval-based distribution
operations (IBDO). This technique discretizes input distribution curves and
hence can handle arbitrary distributions. Further, this technique avoids
approximations by using discretizations that provide an envelope around the
input curve being discretized, thereby using a strategy of bounding instead of
approximating. Outputs are also bounding envelopes. In this technique, inputs
for which the dependency relationship is unknown lead to widened output
envelopes that incorporate the additional uncertainty this poses while retaining
the philosophy of bounding instead of approximating.
A project website has been established at http://class.ee.iastate.edu/berleant/home/Research/Pdfs/PSERC/index.htm. This website contains papers, technical background information, current investigation documentation and project descriptions. Weekly meetings have been conducted almost continuously. Advances on the application of interval theory to uncertain correlated variables have been achieved. Two papers have been submitted to demonstrate the application of this new algorithm. We have extended ISU contract models with existing contingent strategies for risk management. We have augmented the code for system model to examine significant uncertainties at transmission and distribution levels as well as at generation and demand levels. We are preparing conference and journal papers to document these tasks.
The issue of excess width in the envelopes, due to multiple occurrences of variables in the function that combines two random variables, has been solved for monotonic functions. This enables solving the economic dispatch problem under newly lenient assumptions, further closing the books on this important problem. Secondly, the tool has been upgraded to allow min and max operations, enabling solution of concurrent task problems and two-component reliability problems. Thirdly, Pearson correlation has been incorporated into the linear programming constraint structure, enabling both values and ranges of values of correlation to be used. This is an important form of information about partial dependency. We are applying this to Value at Risk problems.
Work progress since the report for the December 2001 IAB
meeting
A project website has been established at pserc.ee.iastate.edu. This website contains papers, technical background information, current investigation documentation and project descriptions. Weekly meetings have been conducted almost continuously. Advances on the application of interval theory to uncertain correlated variables have been achieved. Two additional papers have been submitted to demonstrate the application of this new algorithm to decision analysis and real options. We have extended general contract models with dominant strategies for auction play. We have augmented the code for system model to examine significant uncertainties of equipment availability of transmission, distribution, and generation. We are preparing conference and journal papers to document these achievements.
The comparison of two techniques: excess width in the envelopes, due to multiple occurrences of variables in the function that combines two random variables, and fuzzy linear programming are on-going. This enables solving the scheduling and dispatch problem under newly lenient assumptions, further closing the gaps on this important problem. We are applying this to Real Option Valuation and Value at Risk cases.
Description of work activities and anticipated project
outcomes/deliverables by each project team member during next reporting period
The Berleant and Sheblé project team have conducted joint meetings with the students involved to establish the technical merits of the approaches and to develop the case studies to exemplify applications. Two students have been primarily involved with the advances in algorithms. Three students are involved with the generation of the case studies. One undergraduate has joined the team and will become a graduate student on this project as of this May.
Description of and reasons for any revisions to the
workplan that was reported for the December 2001 IAB Meeting (please
incorporate revisions in above text)
Work tasks have been
altered to reflect the late start of this project and the availability of
students. This project started 6 months
later than expected due to contract delays.
Additional funding for students in this area was subsequently obtained
to complement the work in this project.
Project tasks have not been changed.
Students working on the project during the next reporting
period (names and email addresses)
Andrew Brown – ajbrown@iastate.edu
Chin-Chuen Teoh - ccteoh@iastate.edu (previously supported by National Science Foundation Contract)
Dave Doty - ddoty@iastate.edu (previously supported by National Science Foundation Contract)
Guillermo Gutierrez - ggutier@iastate.edu (supported by grant from Mexico)
Wang Yu - wangy@iastate.edu (supported by EPRC/MidAmerican)
Mei-Peng Cheong - mpcheong@iastate.edu (ISU undergraduate)
Zhang Zianzhong -
zjz@iastate.edu
D. Berleant, J. Zhang, R. Hu, and G. Sheblé, Economic dispatch: applying the interval-based distribution envelope algorithm to an electric power problem, SIAM Workshop on Validated Computing 2002, Extended Abstracts, Toronto, May 23-25, pp. 32-35. (Refereed.)
D. Berleant, L. Xie, J. Zhang, and G. Sheblé, An improved tool for distribution envelope determination, a technique for interval-based, verified arithmetic on random variables, SIAM Workshop on Validated Computing 2002, Extended Abstracts, Toronto, May 23-25, pp. 26-31. (Refereed.)
G. Sheblé and D. Berleant, Bounding the composite value at risk for energy service company operation with DEnv, an interval-based algorithm, SIAM Workshop on Validated Computing 2002, Extended Abstracts, Toronto, May 23-25, pp. 166-171. (Refereed.)
Daniel Berleant, Jianzhong Zhang, and Gerald Sheblé, Using correlation to improve envelopes around derived distributions, draft to be submitted.
Daniel Berleant, Jianzhong Zhang, and Gerald Sheblé, On bounding times to failure of two-component systems, submitted 4/02.
Helen Regan, Scott
Ferson, and Daniel Berleant, Equivalence of five methods for bounding
uncertainty, accepted to International Journal of Approximate Reasoning pending
revision.
Daniel Berleant, Lizhi Xie, Jianzhone Zhang and Gerry Sheble. “On Completion Times of Networks of Concurrent and Sequential Tasks.” Under revision. (Available in the 2001 Publications folder on the PSERC website.)
Click here to return to the Project Overviews