white paper i research and development report

33
MISO Enhanced Combined Cycle White Paper I Research and Development Report January 31, 2018 Issue ID: MR002

Upload: others

Post on 18-Oct-2021

4 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: White Paper I Research and Development Report

MISO Enhanced Combined Cycle

White Paper I

Research and Development Report

January 31, 2018

Issue ID: MR002

Page 2: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

1

1. Executive Summary

In 2017, MISO performed a benefit analysis and computational study based on the

combined cycle data collected from stakeholders with MISO’s R&D-developed prototype

software that included Enhanced Combined Cycle (ECC) modeling. The study estimated about

$14 to $34 million annual production cost savings from the enhanced modeling. The

computational study based on sample production Day-Ahead Security Constrained Unit

Commitment (SCUC) also shows acceptable performance results. Therefore, MISO

recommended moving forward with conceptual design and implementation. Meanwhile, MISO

continues to work with research partners to further improve ECC modeling and solution

performance.

MISO formed the Enhanced Combined Cycle Task Team (ECCTT) in September 2017 to

facilitate open and transparent dialogue with stakeholders to discuss and develop the best

design options for enhanced combined cycle modeling. MISO has hosted monthly ECCTT

meeting with stakeholders since October 2017. The meetings have informed stakeholders on

configuration and modeling options and obtained consensus on design choices. MISO plans to

continue the conceptual design discussion in 2018.

Since the start of the energy market in 2005, MISO has continuously enhanced its market

design and market clearing system to deliver greater value to its region. MISO established the

market roadmap process as a transparent method to prioritize future market enhancements.

Within its market clearing process, MISO solves one of the largest and most complicated

SCUC problems among all Regional Transmission Organizations (RTOs). The introduction of

new features in energy and ancillary services markets can add computational complexity to

Page 3: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

2

existing market clearing problems. Enhanced combined cycle modeling has been a high-

priority market roadmap initiative for several years given the potential benefits for better

utilizing the various configurations of these flexible and complex resources. It can also reduce

the risk of combined cycle owners and incentivize them to offer true cost. MISO stakeholders

have shown great interest in this enhancement since 2011. However, initial computational

study indicated significant performance challenges. Starting in 2013, MISO dedicated internal

R&D staff and collaborated with vendors, research partners and MISO operations, to improve

market system performance, resulting in a reduction in its Day-Ahead market-clearing window

from 4 to 3 hours. This achievement enabled MISO to better align its energy markets with

natural gas markets and meet scheduling requirements in FERC Order 809. Improved SCUC

formulation and solution approaches developed from this research made possible enhanced

combined cycle modeling in MISO’s current market systems.

This whitepaper summarizes the R&D progress, the performance and benefit analysis, unit

commitment modeling options and MISO recommendations based on stakeholder feedback. It

also documents the main conclusions, design choices on key unit commitment modeling, and

contains a list of the key issues to be addressed in the conceptual design and implementation

phases throughout the remainder of the project.

Page 4: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

3

2. Background

A combined cycle power plant is a combination of gas and steam turbine units where the

waste heat from the gas turbine(s) is used by steam turbine(s) to generate additional power.

Among current thermal generation options, combined cycle gas turbines (CCGT) have a lower

levelized cost of electricity, higher efficiencies, lower CO2 emissions, better operational

flexibility and faster response. Therefore, there is an upward trend of installing combined cycle

units and CCGTs are expected to become a larger portion MISO’s generating portfolio, based

on planning studies.

MISO evaluation of configuration-based CCGT modeling began in 2011. At first, the poor

initial run-time performance of the prototype market-clearing software with a CCGT

configuration model prevented further design and implementation steps. Recent improvements

in run-time performance now allow for the next stages of development.

With a configuration-based CCGT model, each configuration is essentially modeled as a

generator and a transition matrix represents the relationship between the configurations. Due

to the increased integration of renewable energy resources, CCGTs are becoming increasingly

critical to system operations due to their fast-ramping capability, quick response and relatively

low cost. Better modelling of CCGTs in the market-clearing engine can result in production

cost savings and provide participants with the ability to more accurately model their true costs.

Researchers have included CCGTs in several real-world SCUC implementations, using a

variety of modeling approaches. There are four main approaches used to represent the

CCGTs.

1) Aggregate model. This model simplifies a CCGT as a pseudo conventional unit by

Page 5: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

4

allowing the combine-cycle owner to provide an aggregate offer. This model sacrifices the

accuracy of CCGTs because the aggregate offer in a given interval cannot accurately

represent the operating cost of multiple available configurations of the individual gas

combustion turbines (CT) and steam turbines (ST) that make up the aggregate CCGTs and the

costs to transition between them. Several RTOs, including MISO, are using this model.

2) Separate modeling. Each CCGT component (sub-unit) is modelled separately and the

production costs of the individual units are reflected in the objective function. This model

preserves the generation characteristic for each sub-unit but it does not reflect the relationship

between the ST and CT(s). MISO currently allows combined cycle plant owners to offer the

plant either as an aggregate combined cycle (CC) or as individual CTs. This model is

particularly effective when the steam turbine is out of service and the CTs can be run

independently in a simple cycle mode. Some participants at MISO also split one ST into

multiple partial STs and model each “CT-partial ST” as one aggregate CC. This model can

allow a certain level of offer flexibility but each “CT-partial ST” pseudo aggregate may run into

difficulty following instructions.

3) Configuration-based CCGT model. A configuration is a pre-defined set of CTs and/or

STs. This model can better represent the operational modes, transition characteristics and

operating cost of CCGTs. However, it does not directly consider the physical units comprising

the configuration, and thus may not be able to accurately reflect operating restrictions like

minimum-up or minimum-down time constraints on each physical component.

4) Individual component-based CCGT model. This model describes the operating

constraints for each sub-unit (each ST, CT, duct burner, etc.) with a graph to represent the

relationship between valid configurations and physical components. However, this model may

Page 6: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

5

incur significant computational complexities beyond the configuration-based CCGT model. As

the number of CCGT units increase, there is a larger-and-larger impact on SCUC

computational performance. This model is still in the research stage.

Page 7: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

6

3. Recent Breakthrough

ERCOT, CAISO and SPP implemented configuration-based combined cycle models in their

market clearing processes. Some of them started with limiting the total number of

configurations to reduce the risk to market system performance. Generally, MISO has a larger

number of total generators and/or a larger number of combined cycles than these

organizations. The prototype software performance did not demonstrate acceptable

performance for production implementation until recent performance advancements in the

SCUC formulation [1]. These advancements resulted from MISO’s dedicated efforts to improve

its production Day-Ahead market software run-time performance, which successfully reduced

the Day-Ahead clearing time in 2016 from 4 hours to 3 hours [2] [3], and better-aligned the

natural gas and electricity markets. These enhancements resulted in a 30 percent

improvement (reduction) in SCUC Mixed Integer Programming (MIP) performance, a 37

percent improvement in Day-Ahead software solving time, and a 66 percent improvement in

the configuration-based combined cycle prototype Day-Ahead SCUC MIP performance.

In 2017, MISO evaluated the impact of configuration-based CCGTs on market management

system performance. A test case with 96 configuration-based CCGTs (more than double the

existing 44 CCGTs in the current MISO market) resulted in 13% increase over the Aggregate

model in the Day-Ahead SCUC MIP solving time. The satisfactory performance results have

led to the decision to proceed with further product efforts. Based on the recommendation from

the R&D study, MISO will move ahead with an enhanced combined cycle design in 2018 and

anticipated implementation in 2020.

MISO R&D continues to work with research partners to address the limitations of the

Page 8: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

7

configuration-based model. Other RTOs have developed complicated ways to work around

modeling and computational limitations. MISO has further developed a prototype “hybrid

configuration and individual component model” to combine the benefit of configuration-based

models and individual component models [4].

Throughout this document, the term “enhanced combined cycle modeling” is used to refer

to configuration-based and/or “hybrid configuration and individual component-based” CCGT

modeling.

The open and transparent dialogue with stakeholders provided in the ECCTT monthly

meetings has enabled MISO to narrow down the modeling options and provide important

inputs to commence conceptual design and software design.

Page 9: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

8

4. Computational Study and Benefit Analysis

MISO collected combined cycle data from participants in 2012 and 2013 to develop an

R&D prototype for configuration-based combined cycle modeling and to estimate the benefit of

the enhanced model. MISO utilized the data from several complex market cases during the

January 2014 polar vortex period to develop and test the prototype. The initial performance

was unacceptable. However, with the continued R&D effort between 2014 and 2016, MISO

successfully achieved performance to meet production requirements [1] (Table 1).

Table 1: Comparison on performance on one complex case with enhanced combined cycle modeling

With the increase in MISO CCGT resources following integration of the South Region at

the end of 2013, and the sustained decrease in natural gas prices, the data collected in 2012

and 2013 did not align well with resource offers in current production cases. Thus, MISO

requested updated data from participants in March 2017 to test the latest R&D prototype

software and perform a benefit analysis.

There are currently 51 combined cycles providing some 29 GW of capacity. Some

participants model one combined cycle as multiple aggregates: a portion of a single ST

associated with each CT to create an independently modeled CT-partial ST aggregates. Some

other plants only run with one online configuration and do not require a configuration based

model. Several combined cycle owners did not respond to the data request. It is unclear

Page 10: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

9

whether they plan to use the configuration-based model in the future. Eventually, 44 resources

in the current production model were converted to 31 configuration-based ECC models using

the collected data. There are 122 configurations under the 31 ECC plants.

MISO randomly selected one Day-Ahead market case from each month (including

weekdays and weekends) from January to June 2017. For this performance analysis, the

prototype software used a configuration-based combined cycle model based on the

formulation in [1] where all unit commitment constraints are on configuration level (as

described for the configuration-based model in Chapter 2: Background). It does not reflect

recent development of the hybrid configuration and individual component model (described in

Chapter 2: Background and a research paper [4]) to allow certain constraints to be enforced on

the components as shown in section 5.1.2.

MISO explored the use of two of the best commercial Mixed Integer Programming (MIP)

solvers, CPLEX 12.6 and Gurobi 7.5, under both the aggregate model and the prototype

configuration-based combined cycle model. Currently MISO uses CPLEX in its production

market systems, but MISO has also been working with Gurobi on R&D collaboration. The

testing and studies reveal (Table 2):

• New configuration-based combined cycle performance under Gurobi 7.5 is comparable

to current production performance with the current production aggregate combined

cycle model.

• A future scenario with 96 enhanced combined cycles representing 49.5 GW of capacity

was also studied. Day-ahead SCUC MIP solving time increased 13% with CPLEX and

was considered acceptable.

Page 11: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

10

Some RTOs include both CPLEX and Gurobi solvers in their production market clearing

process to provide backup in case one solver does not perform well with a particular problem.

Given the performance of Gurobi, the MISO R&D group suggests incorporating Gurobi solver

into the market-clearing engine, especially for the Day-Ahead and Forward Reliability

Assessment Commitment (FRAC) SCUC.

Table 2: Computational Performance Summary

Test Machine Specifications: CPU: E5-2690 @ 2.90GHz; System: Windows 7 64-bit, RAM: 64.00 GB

The same data collected in March 2017 was also used to evaluate the benefit of the

configuration-based combined cycle model. The data reflects the March natural gas prices.

Benefit numbers were derived in the following manner:

1) Re-run Day-Ahead SCUC under production data (excluding virtuals). The production

cost from the each market day was used as the base cost.

2) Replace the aggregate CC with the configuration based ECC data. Re-run Day-

Ahead SCUC (excluding virtuals). The production cost from the market day was

used as the new cost.

CPLEX GUROBI CPLEX GUROBI CPLEX GUROBI

Time (s) Time (s) Time (s) Time (s) Time (s) Time (s)

1/20/2017 328 296 399 280 460 253

2/9/2017 311 223 573 505 554 330

3/13/2017 471 209 566 318 551 351

4/22/2017 333 146 397 203 403 192

5/22/2017 797 1211 972 624 932 639

6/8/2017 331 224 644 310 647 321

Average 429 385 592 373 591 348

New Configuration

based (cold start)

New Configuration based

(from initial commitment)

Existing Aggregate

(cold start)Case

Page 12: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

11

3) The daily benefit value is the production cost saving, i.e., the difference between the

base cost and the new cost.

MISO performed several analyses to ensure the results were reasonable.

First, MISO carefully reviewed the gas prices and existing aggregate CC offers in

production cases to avoid overstating the benefit. The January case benefit was much higher

than other cases. The analysis identified that January offers were historically higher than

March offers, which also held true in 2017. Replacing aggregate CC offers in the January case

(corresponding to the January gas price) with the collected offers (corresponding to March gas

price) could overstate the benefit. Hence the January case was excluded in the benefit

estimation to avoid introducing errors from gas price differences.

Second, MISO mapped the existing aggregate CC outage data to the new

configurations representing the same resource in order to properly reflect the outage status

and initial on/off status. The initial on/off status of ECC configurations were derived based on

the production aggregate CC and individual CT/ST initial on/off times.

Third, MISO decided to exclude virtuals in this benefit analysis because virtual offers

and bids are based on existing Day-Ahead and RT prices. Virtual offers and bids, which are

financial-only transactions in the Day-Ahead market, can cause an overstatement of the

benefit from virtual production cost reduction due to overall lower prices under a configuration-

based model.

The estimated annual savings are $14 million to $34 million, based on the smallest daily

savings and the average daily savings (Figure 1).

Page 13: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

12

Figure 1: Production cost savings for ECC model addition

In summary, these are the conclusions of the performance and benefit analysis:

• The benefit study indicated a potential $14 million to $34 million annual savings.

• Performance studies on current and future Day-Ahead SCUC cases are acceptable.

• Suggest incorporating the Gurobi solver in addition to the current CPLEX solver.

• Recommend moving forward with conceptual design.

$-

$20,000

$40,000

$60,000

$80,000

$100,000

$120,000

$140,000

$160,000

$180,000

9-Feb 13-Mar 22-Apr 22-May 8-Jun

Pro

du

cti

on

co

st

sav

ing

s (

$/d

ay)

Page 14: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

13

5. SCUC Modeling Recommendation

Through its research, study and discussion with stakeholders at ECCTT, MISO reaches

conclusions on key modeling options and summarizes the Enhanced Combined Cycle

modeling recommendations in this section. MISO also provides examples to illustrate the

implications.

The configuration-based model allows unique operational parameters such as energy offer

curves and ramp rates, to be offered for each configuration. However, certain constraints, such

as minimum-run and minimum-down times, are driven by physical reality of the individual

components (CT, ST, duct burners, etc.). Modeling these parameters on the configuration level

may cause a physical infeasibility. Some RTOs have introduced group constraints as an

attempt to address this issue. This approach requires mapping between the configurations and

groups, based on linking components that drive the constraints. It may also require additional

validations to prevent conflicts between the inputs of the configurations and the groups.

MISO did further research with academia and developed hybrid models to allow

constraints to be modeled either on the configuration level or on the component level [4].

Enforcing constraints, such as minimum-run time and minimum-down time, directly on the

physical components that drive these constraints is more straight-forward and easier to

maintain. MISO has discussed this with stakeholders at ECCTT meetings, and has collected

their feedback through multiple questionnaires.

Page 15: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

14

5.1 Modeling of the Key Unit Commitment Constraints

5.1.1 Configuration Constraints

The general consensus is that the following constraints should be modeled on the

configuration.

• Incremental energy offer

• Reserve offers

• Ramp rate

• Maximum and minimum output limits, dispatch and commitment status

• No-load cost

• Regulation/spinning reserve qualification

• Maximum daily energy

• Maximum daily start

5.1.2 Component Constraints

MISO determined that these constraints are best suited for modeling on the component

level:

• Minimum run time

• Minimum down time

• Maximum run time

MISO had several rounds of stakeholder discussions on these minimum-run time and

minimum-down time constraints. Since these constraints are mainly driven by individual

components, the accuracy of this modeling approach was selected over applying these

parameters to configurations. If physical components have different minimum-run or down

Page 16: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

15

times, the offers on the configurations may be too conservative (i.e. inefficient) or lead to

infeasible operating schedules.

As an alternative, consideration was given to alleviating the configuration minimum run and

down time challenges by enforcing the following rules (i.e., “configuration-enhanced minimum

run/down”):

• For an up transition, do not enforce the minimum-run time on the “from

configuration”

• For a down transition, do not enforce the minimum-down time on the “to

configuration”

However, the configuration model does not include individual component on and off times

for full consideration. This limitation can potentially result in schedules that cannot be realized

without violating an individual component minimum-run time or minimum-down time constraint.

Enforcing the minimum-run and down time on an individual component fundamentally resolves

such issues. Appendix I has an example that compares the options of enforcing minimum-run

time and minimum-down time constraints on the configuration model, the configuration-

enhanced model and the individual component model. Both the configuration model and the

configuration-enhanced model can result in infeasibility. For this reason, these constraints are

proposed to be modeled on the individual component level, instead of on the configuration.

Recent R&D efforts in innovative and effective mathematic modeling allow unit commitment

constraints to be enforced both on the configuration and the component levels simultaneously

[4].

Page 17: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

16

5.1.3 Constraints Between Configurations

The transition from one configuration to another involves the startup and/or shut down of

one or more components. In this document, a configuration “startup” refers to the plant

moving from the “AllOff” configuration to an “on” configuration, while a configuration

“transition” occurs between two different “on” configurations. Two types of constraints must be

modeled for the startup/transition:

1) Startup time/startup cost, or transition time/transition cost

2) Startup time and notification time, or transition time and transition notification

times, for the first startup/transition in SCUC.

Similarly, these constraints could be enforced either on the configuration level or on the

individual component level. Modeling on the component level may bring more accuracy on the

cost in relation to the down time of the component. However, the cost of starting the ST is

mainly driven by the CT fuel needed to heat the ST. Thus, the derivation of individual

component startup costs may not be straightforward for the ST. Given that, MISO proposes

modeling the startup/transition constraints at the configuration level.

Within the configuration level, there are several options for how to model startup/transition

time and cost. Most participants agreed on a mixed model for each pair of “from configuration”

to “to configuration” transitions:

• Three configuration startup times associated with the state (i.e. hot, cold, intermediate)

of the configuration and costs for starting the CCGT from an offline state (which is called

the AllOff configuration in this document) (Figure 2).

Page 18: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

17

o Hot/intermediate/cold configuration startup costs associated with

hot/intermediate/cold time for the “to configuration” (see the 2X1 startup in Figure

2)

o Note: the time between the new 2X1 start and the last 2X1 off (i.e., ������

�����

) is compared to the three startup times to determine the cost. For this

specific example, the configuration-level modeling may introduce inaccuracy

since there are other online configurations in between ����� and ����

��. The

state and startup costs of individual CTs and the ST may be different from the

one determined based on the 2X1 configuration. Given the complexity and lack

of experience with deriving the individual component startup costs, MISO

proposes to begin with this startup model and decide whether it needs to be

improved after gaining more experience.

Figure 2: Three startup times and costs when starting from AllOff

• Single transition time and cost between specified configurations (if not transitioning from

AllOff) (Figure 3).

Page 19: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

18

o Single transition cost (“CT2+ST” �2X1 in this example)

o Transition time is used to disqualify reserves at the end of the “from transition.”

o Transition time is one factor impacting the on time of “from configuration,” i.e.,

“transition time”≤ “from configuration on time”

o It is assumed that energy is dispatchable during transition. In real time,

participants can send a control mode status (CMOD) of 3 to the Automatic

Generation Control (AGC) if it cannot follow the energy dispatch instructions

during the transition. AGC will simply echo back the current MW to the resource

as the set point under CMOD=3.

Figure 3: Single transition time and cost when not transitioning from AllOff

• Three configuration notification times (hourly parameter used in conjunction with the

corresponding startup time for the first start up in SCUC when initially at AllOff)

• Single transition notification time (used in conjunction with transition time for the first

transition in SCUC when initially not AllOff) (Figure 4)

Page 20: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

19

Figure 4: First transition/first startup

5.2 Examples of the Recommended Options

MISO has developed examples to illustrate the enhanced combined cycle model under this

proposal. These examples assume a combined cycle plant with CT1, CT2, CT3, ST and Duct

Burner (DB).

5.2.1 Registration

In registration, participants will register the most commonly used plant configurations.

MISO proposes to start with a maximum of seven configurations (including AllOff), although

this number may be adjusted based on further computational studies. Multiple equivalent

configurations can be represented by one, such as one CT and one ST. In this example,

assume the combined cycle plant is registered as follows (Figure 5):

Page 21: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

20

Figure 5: Configuration registration example

In this registration example, there is one registered 2X1 configuration, but physically,

the plant has three combinations of ways that it could operate in 2X1: “CT1, CT2, ST”; “CT1,

CT3, ST”; and “CT2, CT3, ST”. Substitution of equivalent components, CTs in this example,

are allowed. For enforcement of offered operational constraints, MISO will assume the

resource is running with the equipment specified during registration. If the resource participant

chooses to deviate from this assumption, the plant operator is responsible for substituting

resources in a way that avoids conflicts between the MISO commitment plan and component

operating constraints as further explained in Section 5.2.3: CT Substitution. MISO will receive

ICCP telemetry for CTs and STs, and will check that the ICCP configuration code matches the

code registered for the committed configuration, but MISO will not check that specific

components are running when evaluating the overall output of the plant for measurement and

verification.

5.2.2 Transition Matrix

During registration, market participants also need to provide the transition matrix

describing the possible transitions between registered configurations (Figure 6). For each valid

transition not from AllOff, the parameters “transition time/transition notification time/transition

cost” will need to be specified in the resource offer.

Page 22: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

21

Figure 6: Enhanced combined cycle transition matrix

Some RTOs require the transition from AllOff to be valid for all online configurations,

because it may help reduce the computational impact of the enhanced combined cycle model.

Also, it may reduce the impact introduced by inaccuracy from hourly intervals in Day-Ahead

and any Reliability Assessment Commitment. For example, assume “AllOff�1X1-DB” is

operationally an invalid transition. Day ahead may commit from AllOff �1X1�1X1-DB. Even if

the plant is required to be in 1X1 for just 10 minutes, day ahead will commit 1X1 for at least 1

hour before it can transition to 1X1-DB. The MISO design does not preclude the resource

participant from offering an “AllOff�1X1-DB” transition if it is deemed appropriate for the

participant to embed the transition through the 1X1 configuration as part of the 1X1-DB startup

time and cost. MISO has not seen the computational need to enforce the requirement that

transitions between AllOff and all registered online configurations need to be valid. MISO will

perform more case studies to evaluate if it is necessary to enforce such a rule. Most

participants consider this rule acceptable based on questionnaire feedback.

Page 23: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

22

5.2.3 CT Substitution

Participants are very interested in the ability to substitute equivalent CTs. First of all,

with a maximum of seven allowed configurations, it may not be feasible to register all the

possible configurations. In the example shown in Section 5.2.1: Registration, only one of the

three possible 2X1 configurations is registered. Secondly, there might be some contract

requirement or constraints not modeled in MISO SCUC, e.g., balance the number of starts or

online hours of the CTs at the same plant.

Operationally, if the CT associated with the committed configuration is unavailable, it will

provide a reliability benefit if another equivalent CT can substitute and meet the MISO

schedule. For these valid reasons, MISO proposes:

1) Performance measurement will continue to be on the total combined cycle plant level

2) Individual unit substitution will be allowed

• MISO will not change the configuration in the commitment plan for component

substitutions unless there is valid reason for operators to override the

commitment plan’s configuration.

• Participants must properly manage the CT substitution to avoid violating any

individual physical component constraints.

3) With the proposed maximum of seven configurations, participants may choose to use a

subset of possible configurations to represent other equivalent configurations. The

representative configuration will be mapped to the representative CT inside the MISO

clearing software, based on the participant’s registration data.

4) Participants can start a different equivalent CT as long as it provides the scheduled

capacity and meets other operational requirements.

Page 24: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

23

• MISO will assume the plant is operating according to the configuration committed

by MISO.

• MISO will not account for any inconsistency caused by swapping CTs that

deviate from the registered configuration initiated by participants.

• Participants need to make sure to maintain the consistency and ensure

feasibility.

In the example in Figure 7, assume MISO Day-Ahead commitment is as shown on

“committed configuration,” i.e., 1X1-A, 2X1, 2X1-DB, 2X1 and 1X1-B. They are mapped to

physical components based on the registration (Table 5). MISO expects CT1 and/or CT2 to be

online.

For the purpose of enforcing minimum-run time and minimum-down time constraints on

individual components, the on/off times of individual components are derived based on the

on/off time of the configurations in the MISO commitment plan. Consider the example in Figure

7. At the end of the day, the on-time of CT1 is -2 hours (i.e., off 2 hours) and the on-time of

CT2 is 18 hours. The on/off times will be important for determining the next day commitment.

The participant can choose to start CT2 and CT3 instead. In the example, assume the

participant uses CT2 to substitute CT1 and CT3 to substitute CT2. What happens in reality is

that CT2 is off for two hours and CT3 is on for 18 hours at the end of the day. As long as the

participant can maintain the consistent mapping (CT1�CT2, CT2�CT3), the commitment

from MISO SCUC will be feasible. In this example, the unit commitment constraints are

enforced on configurations or components associated with CT1 and CT2. The participant

elected to swap components so the constraints actually represent CT2 and CT3, respectively.

Page 25: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

24

However, if the consistent mapping is not maintained, participants will bear the risk of

receiving an infeasible commitment. For example, if at hour 23, the participant transitions to

CT1+ST, the actual initial status will become CT1: 2h, CT2: -2h, CT3: -2h. MISO will consider

the committed 1X1-B (“CT2+ST”) is online for the last two hours. Inside the next day SCUC, it

still considers the initial status as CT1: -2 hours and CT2: 18 hours at the beginning of the first

hour. The participant’s CT substitution may result in an infeasible commitment.

Figure 7: Example swapping on combined cycle components

5.2.4 Treatment of Outage Information Inside Market Clearing Software

Since substitution is allowed, it will not be appropriate to use CROW outage information to

override the commitment status of the configurations. For example, if only CT1 is outaged in

CROW, treating 1X1-A as outaged in SCUC may be inappropriate since participants may use

CT3 to substitute CT1.

Meanwhile, outage information is useful for multi-day studies when accurate offer data is

not available beyond the operating day. Hence, MISO proposes that if all the physical units

under a combined cycle plant are outaged in CROW, then all the registered configurations

Page 26: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

25

under the combined cycle are also considered as outaged in the clearing software. If only a

subset of the physical units is outaged in CROW, then the CROW outage information is

ignored and the offered commitment status is used in the clearing software. This is similar to

MISO’s treatment of aggregate combined cycle resources today where the aggregate

combined cycle resource is only modified (considered unavailable) if all components of the

aggregate are outaged in CROW.

5.2.5 Market Offers

This section illustrates the structure of the key unit commitment parameters.

1) “Hot to cold time,” “hot to intermediate time” and three startup costs are offered on

each valid configuration that can be transitioned from AllOff. They are daily

parameters and are used to enforce three startup times and costs when the

configuration starts from AllOff (Table 3).

Table 3: Configuration startup cost and time

2) Three startup times and notification times can be offered for the first transition when

the initial configuration is AllOff. They are hourly parameters (Table 4).

Configuration name COLDSTARTUPCOST INTERSTARTUPCOST HOTSTARTUPCOST HOTTOCOLDTIME(h) HOTTOINTERTIME(h)

1X1-A 1500 1000 500 10 4

1X1-B 1500 1000 500 10 4

2X1 2500 2000 700 12 6

2X1-DB

3X1 3500 3000 900 12 6

3X1-DB

AllOff

when start from alloff

N/A (invalid to start from AllOff)

N/A (invalid to start from AllOff)

Page 27: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

26

Table 4: Configuration startup and notification time

3) Minimum-run time, minimum-down time and maximum-run time are offered on

individual components (Table 5).

Table 5 Individual component parameters

4) Other parameters are offered and enforced at the configuration level:

Daily parameters include:

• Maximum daily energy

• Maximum daily start

• Reserve qualifications

Hourly parameters include:

• Incremental energy offer (multi-segment offers similar to other resources)

• Reserve offers

• Ramp rate

• Maximum and minimum output limits

Configuration nameHOUR(dd-mm-yyyy

hh)

COLDSTARTUP

TIME(h)

INTERSTARTU

PTIME(h)

HOTSTARTUP

TIME(h)

COLDNOTIFICATION

TIME(h)

INTERNOTIFICAT

IONTIME(h)

HOTNOTIFICATIO

NTIME(h)

1X1-A 15-JAN-2014 01 3 2 1 3 1 1

1X1-B 15-JAN-2014 01 3 2 1 3 1 1

2X1 15-JAN-2014 01 3 2 1 3 1 1

2X1-DB 15-JAN-2014 01 3 2 1 3 1 1

3X1 15-JAN-2014 01 3 2 1 3 1 1

3X1-DB 15-JAN-2014 01 3 2 1 3 1 1

AllOff 15-JAN-2014 01 3 2 1 3 1 1

15-JAN-2014 02

15-JAN-2014 03

INDIVIDUAL UNIT NAME MINDOWNTIME(h) MINUPTIME(h) MAXRUNTIME(h)

CT1 8 5 N/A

CT2 8 5 N/A

CT3 8 5 N/A

ST 12 10 N/A

DB 2 2 N/A

Page 28: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

27

• Dispatch and commitment status

• No-load cost

5.2.6 Next-Step Computation Study

The stakeholder discussions at the ECCTT have focused on providing a better

understanding of the proposed ECC model. Some of the elements were not in the 2017

prototype used for the analysis. MISO recently requested a new set of data including

suggested individual component parameters. MISO will align the prototype software with this

proposal and perform further computational studies and inform stakeholders and the

development team if any potential issues are identified.

Page 29: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

28

6. Key Design Areas

MISO has started the conceptual design discussion at ECCTT. The following key areas

have been identified. Table 6 shows the complexity and current progress of each area. MISO

will continue working with stakeholders to reach consensus on major design issues by June 1,

2018. Draft modifications to the MISO Tariff will be completed by December 31, 2018.

Table 6: Key conceptual design issues (as of 1/31/2018)

Area Key issues Complexity Progress

Registration and Modelling

Offer Structure High Finalizing

Number of Configurations Medium In progress

Measurement and Verification

Telemetry Low In queue

Outage management Low Start

Market Clearing (Commitment and Dispatch)

Commitment in day ahead and RT

High In progress

Transition – Reserve modeling

High In progress

Transition – Energy Dispatch

Low In progress

Settlement DA RSG and RT RSG High In progress

DAMAP and RTOSG High Start

Reserve Buyback Medium In queue

Pricing Startup/Transition cost; operating limits

High In queue

Page 30: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

29

7. References

[1] Y. Chen and F. Wang, "MIP Formulation Improvement for Large Scale Security Constrained Unit Commitment

with Configuration based Combined Cycle Modeling," Electric Power System Research, Vol. 148, July 2017.

[2] Y. Chen, A. Casto, F. Wang, Q. Wang, X. Wang and J. Wan, "Improving Large Scale Day-Ahead Security

Constrained Unit Commitment Performance," IEEE Trans. Power Syst, Vol. 31, Issue 6, Nov. 2016.

[3] "Experience and Future R&D on Improving MISODA Market Clearing Software Performance,"

https://www.ferc.gov/CalendarFiles/20170623123549-M1_Chen.pdf, June 2017.

[4] C. Dai, Y. Chen, F. Wang, J. Wan and L. Wu, "A Tight Configuration-Component Based Hybrid Model for

Combined-Cycle Units in MISO Day-Ahead Market," https://arxiv.org/abs/1708.06413, Aug. 2017.

Page 31: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

30

Appendix I: Minimum-Run Time and Minimum-

Down Time Constraints

This section compares three options to model minimum-run time and minimum-down time

constraints:

1) All on the configuration level based on minimum-run and minimum-down offers on the

configuration level

2) Configuration enhanced model

• For up transition, do not enforce minimum run time on the “from configuration”

• For down transition, do not enforce minimum down time on the “to configuration”

3) Enforcing the minimum run/down time on particular individual component

Assume a combined cycle plant has five physical units: CT1, CT2, ST and Duct Burner

(DB). The individual component minimum-run time and minimum-down time parameters (Table

7) are shown as:

Table 7: Individual component parameters

hours Min Run Min Down

CT 4 8

ST 10 12

DB 2 2

The first scenario examines a transition from 1x1-A to 2x1 followed by a transition back

to 1x1. Figure 8 depicts the scenario and configuration minimum-run and down times that may

be applied to the potential configurations. To prevent infeasibility, if each of these

configurations can start and stop via transition with AllOff, the minimum-run and down times

Page 32: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

31

used at the configuration level must be consistent with the most constraining component, the

ST.

Figure 8: The first example with two transitions

For the first example shown in Figure 8, the constraints enforced by each of the three

models are described in Table 8. For the first transition from 1X1-A to 2X1, the configuration

model appears to require an unnecessarily long time in the 1X1-A configuration prior to the first

transition. For the second transition, both configuration and configuration enhanced models

appear to require an unnecessarily long time prior to allowing the transition from 2X1 to 1X1-A.

Table 8: Comparison of constraints enforced under three different models (first example)

Page 33: White Paper I Research and Development Report

MISO Enhanced Combined Cycle Research and Development Report

32

The second example looks at the potential scheduling constraints around transitions from

1X1-A to 2X1 to 1X1-B. Figure 9 describes the scenario and the scheduling implications of

each model. In this example, all three models require the 1X1-A configuration to be on at least

10 hours before transitioning to AllIOff. Because both the configuration and configuration

enhanced models apply minimum down time to configuration 1X1-B, these models would allow

the 1X1-B configuration to come online prior to the ST component (or the CT) having met its

minimum-down time. The individual model recognizes that the ST went offline with the 1X1-A

configuration and must satisfy the ST minimum-down time before the 1X1-B configuration can

be started (Figure 9).

Figure 9: Second example to compare three models