Parameter Tuning for Uneconomic Adjustments - … · 2018-02-10 · Uneconomic Adjustment Scheduling Run and Pricing Run ... (Self Provision) – which include only MW quantities,

Stakeholder MeetingMay 13, 2008

Parameter Tuning for Uneconomic Adjustments

Lorenzo Kristov, Principal Market Architect

2 Parameter Tuning Stakeholder Meeting May 13, 2008

Topics for Discussion

Objectives of Current Parameter Tuning EffortParameter Tuning Concepts

Uneconomic AdjustmentScheduling Run and Pricing RunPotential for Extreme ResultsLAP Demand Clearing in the IFM

Need for Changes to MRTU TariffIllustrative Examples (Edward Lo)Parameter Tuning Analysis OverviewProposal for Ongoing Updates to ParametersInitial Results for IFM Parameters (Jim Price)Practices of Other ISOs


Objectives of Current Parameter Tuning Effort

Explain role of these “parameters” in the MRTU market optimizations, particularly for Uneconomic Adjustments

Particular focus on a specific uneconomic adjustment case –LAP Demand Clearing

Describe CAISO’s process for “tuning” the parametersParameter settings comprise an essential element of MRTU market design, needed for start-up After initial values are determined, CAISO proposes updating as needed with publication of current parameter parameters

Provide initial parameter settings for ongoing market simulation activitiesDevelop any needed tariff changes for July filing

Compliance regarding LAP Demand ClearingFlexibility to avoid unreasonable, extreme outcomes.


Concepts 1 – Self Schedules and Scheduling Priorities

A fundamental design feature of MRTU since 2002, broadly desired by market participants, is to allow SCs to “Self Schedule” supply and demand at their desired operating levels and thus “opt out” of CAISO market optimizations. Therefore, SC Bid submissions may include:

Economic Bids for Energy and AS – which include both MW quantities and corresponding bid pricesSelf Schedules for Energy, and AS (Self Provision) – which include only MW quantities, no prices.

Related to Self Schedules is a set of Scheduling Priorities –a hierarchical sequence in which different Self Schedule types may be adjusted by the CAISO if necessary to satisfy required constraints (e.g., flow limits, energy balance).


Concepts 2 – Uneconomic Adjustment

MRTU market optimizations first try to solve using only Economic Bids, without adjusting Self Schedules or relaxing any constraints (e.g., transmission limits, AS requirements)

Solution must balance Energy supply and demand, procure requiredAS, and manage congestion to keep scheduled flows within limits

If Economic Bids are insufficient to reach a solution, the optimization resorts to “Uneconomic Adjustments” to reduce Self Schedules or relax some constraints

The frequency of such cases is expected to be low, but will depend on the proportion of Self Schedules in the market optimization.

Incorporating different types of Self Schedules (e.g., TOR, ETC, generic) and other constraints, and assigning each type a different priority for adjustment, requires a system of numerical parameters to enforce the pre-specified priorities.


Concepts 3 – Penalty Prices

Uneconomic Adjustment process is structured by setting “Penalty Prices” – parameters in the optimization that work analogously to bid prices of Economic Bids, but are …

Outside the range of acceptable Economic Bids – outside the Tariff Bid Cap and Bid Floor values, and Set by the CAISO in the software, to enforce the correct adjustment priorities.

Example 1: Penalty price is -$550 in IFM for “price-taker supply bids”(aka “generic” Energy supply self-schedules)

This means that the generic self-schedule will have an associated DEC Bid of -$550 per MW in the optimization, which is outside the allowable range of Economic Bids, and thus forces the optimization to use Economic Bids until it becomes cheaper to use the -$550 DEC Bid.

Example 2: By comparison, penalty price is -$3200 in IFM for Energy supply self-schedules under ETC/CVR

ETC/CVR supply has higher priority protection against curtailment than “generic” supply self schedules.


Concepts 4 – Penalty Price Values

Penalty prices must have extreme values in order to enforce required hierarchies, for example,

Adjust generic self-schedules before adjusting ETC/CVR schedules, and adjust ETC/CVR before TOR schedulesReduce “generic” export self-schedules (not linked to non-RA supply) before reducing internal CAISO demand or export self-schedules linked to non-RA supply=> Penalty prices must be different from Economic Bid range, andsufficiently separated from each other to enforce different priorities

When uneconomic adjustment occurs, the optimization sets prices that are reflective of the penalty price settings – but are not appropriate for settlement.

For example, if the optimization reduces an ETC self-schedule at the penalty price of -$3200 to relieve a transmission constraint, the price differential across the constraint will be at least $3200.


Concepts 5 – Scheduling and Pricing Runs

Problem: Extreme penalty prices needed for enforcing required priorities in Uneconomic Adjustment will result in calculated prices (LMPs, ASMPs) that are too extreme to be used in settlement.Solution: Each MRTU “market process” includes two runs:

First, a “scheduling run” is performed using penalty prices to ensure that Uneconomic Adjustment, if needed, results in the correct resource schedules, Second, a “pricing run” where the penalty prices are replaced by appropriate “pricing parameters” so that LMPs and ASMPs are economically reasonable and consistent with Tariff.

** Market processes include Pre-IFM-CC, Pre-IFM-AC, IFM, RUC, RTUC (aka RTPD), RTD (aka RTID)


Concepts 6 – Simple Example

Suppose there are insufficient Energy Bids offered in the IFM to clear all Self-Scheduled CAISO Demand and Self-Scheduled Exports linked to non-RA supply capacity.Penalty price on reducing such demand is $1600Assume no congestion, no losses – only relevant constraint is the supply-demand balanceScheduling run will:

Reduce such demand until it can be met by the available supply, say to 25,000 MW Determine a scheduling run energy price = $1600

Policy indicates, however, that the price for serving such “price-taker”demand should be set by the Energy Bid Cap, i.e., $500Pricing run will therefore:

Utilize the $500 pricing parameter for a small “epsilon” range around the 25,000 MW demand levelCalculate $500 as the Energy price for settlement purposes.


Concepts 7 – Potential for Extreme Prices Under Economic Clearing

MRTU Tariff says that the optimization will utilize allEconomic Bids before adjusting Self Schedules

The word “all” precludes any flexibility, leading to extreme scheduling and pricing results in certain cases

Example. A large volume of supply self-schedules result in a transmission constraint being overloaded by 2 MW

Only available DEC bid is a 300 MW unit that bid $20 and is 1% effective on the overloaded constraintUnit must be curtailed by 200 MW to obtain 2 MW reduction on theconstrained lineShadow price of constraint will = 200 x $20 = $4000, which means$4000 congestion cost differential across the lineResult does not depend on Uneconomic Adjustment, because Economic Bids were available and were used … BUT …Outcome is not operationally or economically reasonable.


Concepts 8 – LAP Demand Clearing in IFM

Demand in IFM can bid at Default LAP aggregate level, and will be cleared so that resulting price and quantity is a point on the Default LAP demand curve

To ensure this economically consistent outcome, the LDFs for the Default LAP are held fixed in the IFM optimization Unintended consequence – a binding transmission constraint can cause large reduction in cleared demand for the entire Default LAP.

Tariff Section 31.3.1.3 states that in such cases IFM willFirst release self-provided AS to provide energy to serve the load behind the constraint,Second, if needed, relax the transmission constraint Third, if needed, vary some LDFs (not implemented at this time)

Tariff solution is implemented through appropriate penalty price settings.


Some Observations

Generally, extreme scheduling and price outcomes can result from a large proportion of Self Schedules (i.e., small proportion of Economic Bids) in the market optimization.SCs can mitigate their exposure to such extreme outcomes by, for example,

Prioritizing their use of Self SchedulingSubmitting Economic Bids for a portion of the operating range of a self-scheduled resource.

A system of scheduling priorities as included in MRTU requires a system of penalty prices for scheduling purposes and pricing parameters for determining settlement prices. Parameter values must be “tuned” to balance objectives of enforcing priority hierarchy while avoiding outcomes that are operationally or economically unreasonable.


Criteria for Parameter Values

Fulfill MRTU design principle of protecting self-schedules according to the priority order specified in the tariffProvide appropriate economic signals in the form of large magnitude (positive or negative) prices when Uneconomic Adjustments must be utilizedPrevent unreasonable price outcomes, which can occur if there is no flexibility in the hierarchy structure of prioritiesamong the various types of Self Schedule and constraintsAchieve scheduling and pricing outcomes that are consistent with good operational practice and support reliable grid operation.


Need for Tariff Changes

First, the LAP Demand Clearing problem is one case where strict adherence to priorities and enforcement of constraints can lead to extreme outcomes

FERC approved solution of Section 31.3.1.3 but directed CAISO tomodify tariff language to clarify how the penalty price interacts with Economic Bids in this case

Second, the ineffective economic bid of the previous example represents a circumstance where flexibility is appropriate rather than strict adherence to the principle of using all Economic Bids before adjusting Self Schedules

CAISO intends to revise tariff to allow flexibility to avoid solutions that are operationally and economically unreasonable. (see Tariff Section 31.4)


Detailed Illustrative Examples

Presentation by Edward Lo


Scheduling and Pricing Run Illustrative Examples

Formulation and results of scheduling and pricing runsScheduling run establishes energy schedulesPricing run establishes settlement prices

Comparison of using Uneconomic Adjustments to Energy Self Schedules versus Economic Bids to resolve congestion in the scheduling runCalculation of settlement LMPs in the pricing run under Uneconomic AdjustmentRoles of the penalty prices in the scheduling run and pricing run.

Two examples are presented to illustrate:


2 Bus Example – Scheduling Run Setup

Scheduling Run penalty price of -$250 protects the G1 self schedule from curtailment.

NOTE: The value -$250 is convenient for purposes of this example but is not the actual value proposed for use in the MRTU markets.

Transmission line is assumed loseless.

Flow Limit: 80MW

Fixed Load: 200MW

G2G2: [0,150]MW@$50/MWh

G1G1: 120MW Self Schedule[120,200]MW@$10/MWh

120MW 200MW

-$250

$10

Price Curve of G1 in Scheduling Run

1 2


2 Bus Example – Scheduling Run Solution

G1 could have supplied all energy to serve load if no limitation on power flow through line 1-2.Curtailment of G1 self schedule is necessary to resolve the flow constraint violation.Economic signal at bus 1 prior to self schedule curtailment of G1 is $10.

2Flow = 80MW

Fixed Load: 200MW

G2G2 = 120 MWLMP2 = $50

G1G1=80 MWLMP1 = -$250

1

Shadow Price = $300


2 Bus Example – Pricing Run Setup

-$30 Bid Floor price is used for the curtailed G1 self-schedule, with MW lower bound expanded by ε.Epsilon ε is small positive number to:

Ensure feasible solution in pricing runEnsure pricing run MW solution is very close to the scheduling runProvide room for the -$30 price segment of G1 to interact with economic bids of other resources

Flow Limit: 80MW

Fixed Load: 200MW

G2G2: Original Economic Bid

G1

2

200MW

1

G1:

120MW

-$30

$10

Hard Lower Limit at 80-ε MW


2 Bus Example – Pricing Run Solution

MW solution of pricing run is identical to scheduling run.G1 is marginal at 80MW; ε above the lower bound. LMP of bus 1 is set by the -$30 Bid Floor pricing parameter, where this parameter “exceeds on the negative side” the economic signal ($10) prior to self-schedule curtailment.

Flow = 80MW

Fixed Load: 200MW

G2G2 = 120MWLMP2 = $50

G1G1=80 MWLMP1 = -$30

1

Shadow Price = $80


3 Bus Example – Scheduling Run Initial Setup

The -$250 penalty price protects the G1 self schedule from curtailment.NOTE: The value -$250 is convenient for purposes of this example but is not the actual value proposed for use in the MRTU markets.

The three transmission lines are equal in reactance and are assumed loseless.

G2G1

G3

Fixed Load: 500MW

1 2

3

Flow Limit: 260MW

G1: 360MW Self Schedule [0,100] MW @$10/MWh

G3: [0,110] MW @ $70/MWh[110,200] MW @ $500/MWh

G2:


3 Bus Example – Scheduling Run Solution with no Self Schedule Curtailment

Flow of line 1-3 caused by G1 self schedule is 240MW. Flow constraint on 1-3 limits inexpensive G2 to 60MW schedule. Remaining 80 MW supply is from G3.All MW schedules are economical. G2 and G3 are marginal and set LMPs. Shadow price of line 1-3 constraint is $180. For 1 MW increase in line limit, 3 MW of energy supply for fixed load can be shifted from G3 to G2 for reduction of 3*($70-$10) in total bid cost.LMP of -$50 for bus 1 means that 1 MW withdrawal at bus 1 reduces the total bid cost by $50. For 1 MW of such withdrawal supplied from G2 at $10, 1/3 MW of line capacity on 1-3 is freed up. Therefore, 1 MW of energy supply for the fixed load can be shifted from G3 to G2 for $60 bid cost reduction. Hence the net reduction in bid cost is $50.

G2 = 60MW LMP2 = $10

G2G1

G3

Fixed Load: 500MW

1 2

3

Flow = 260MWShadow Price = $180

G1= 360MW LMP1 = -$50

G3 = 80MWLMP3 = $70


Modified 3 Bus Example – Partial Scheduling Run Solution

Congestion is partially reduced without curtailing G1 by shifting energy supply from G2 to G3, to obtain 1/3 MW reduction of flow violation per MW of supply shift. Such re-dispatch is applied up to the MW level shown, which hits the break point in G3’s bid curve. Two options to resolve the last 10 MW of flow violation:

1. Continue the same shift at a high rate of increase in bid cost of 3*($500-$10) = $1470 for 1 MW of reduction in flow violation. LMP at bus 1 would be –($500-2*$10) = -$480. Alternatively,

2. Curtail G1 self-schedule and shift energy supply from G1 to G2 to obtain 1/3 MW reduction of flow violation per MW of supply shift. Rate of increase in bid cost is 3*($250+$10) = $780 for 1 MW reduction in flow violation = lower cost solution.

Consider Flow Limit on 1-3 Reduced to 240MW

G2G1

G3

Fixed Load: 500MW

1 2

3

Flow = 250MW

G1= 360MW

G3 = 110MW

G2 = 30MW


Modified 3 Bus Example – Scheduling Run Solution with Self Schedule Curtailment

Second re-dispatch strategy is adopted for the final scheduling solution.G1 and G2 are marginal. G3 is not marginal.LMP3 = $270. For 1 MW additional withdrawal at bus 3, the lowest cost solution is 2 MW increase of G2 and 1 MW reduction of G1.Without self schedule curtailment, more expensive economic adjustment is necessary to resolve congestion, and LMP1 would be -$480, more negative than the -$250 penalty price of scheduling run.

G2G1

G3

Fixed Load: 500MW

1 2

3

Flow = 240MWShadow Price = $780

G1= 330MW LMP1 = -$250

G3 =110MWLMP3 = $270

G2 = 60MWLMP2 = $10


Modified 3 Bus Example – Pricing Run Solution

Pricing run MW solution is very close to scheduling run.G2 is marginal. G3 moves away from the breakpoint of the bid curve and becomes marginal.G1 is at the lower bound and becomes non-marginal. LMP1 = -$50, the economic signal prior to self-schedule curtailment in the scheduling run set by economic bids of other resources. Should economic signal set by other resources be within the -$30 pricing parameter as in the 2-bus example, pricing run will result in LMP set by this parameter.

G1= 330-ε MW LMP1 = -$50

G2G1

G3

Fixed Load: 500MW

1 2

3

Flow = 240 MWShadow Price = $180

G3 = 80-ε MWLMP3 = $70

G2 = 60+2ε MW LMP2 = $10


Conclusions from the Examples

Penalty Prices for the scheduling run set limits on how extreme (large positive or negative) economic signals can be before adjusting self schedules or relaxing constraints. Thus well-chosen penalty prices can limit the occurrence of extreme scheduling or pricing outcomes due to using ineffective economic bids to resolve constraint violations.Pricing Parameters for the pricing run serve as administrative “floor”(minimum positive or negative magnitude) for the settlement price if the economic signals based on economic bids, i.e., prior to uneconomic adjustment, do not accurately reflect the cost impact of adjusting self schedules or relaxing constraints.In both examples, a high level of self scheduling leads to either

The need to reduce the self schedule to achieve feasibility (2-bus example), orExtreme results based on Economic Bids if all such bids must be used before adjusting Self Schedules (3-bus example)


End of Illustrative Examples Presentation


Parameter Tuning Analysis Overview – 1

Criteria for choosing parameter values –1. Scheduling run should observe required priorities for

protecting Self Schedules and enforcing constraints, consistent with MRTU design and tariff

2. Pricing run should provide appropriate price signals (e.g., shortage of economic bids)

3. Prevent unreasonable extreme price outcomes due to inflexible adherence to curtailment priorities

4. Achieve scheduling and pricing outcomes that are consistent with good operational practice and support reliable grid operation.


Parameter Tuning Analysis Overview – 2

Utilize market simulation cases selected to contain the following conditions –

1. LAP Demand Reduction in IFM/pre-IFMIdentified by release of self-provided AS or relaxation of a transmission constraint

2. Use of Economic Bids with low effectivenessCases having different levels of constraint shadow prices

3. Shortage of supply to achieve system energy balanceFocus on RUC and RTM where demand is set by CAISO forecast

4. Shortage of AS supply to meet requirements5. Export priority enforcement in IFM and HASP.


Ongoing Updating of Parameter Values

Current effort will determine parameter values to be used in market simulation (initial results presented today)CAISO will continue evaluating market simulation cases to verify performance of parameter settings under various market conditions

Will timely inform participants of any updated valuesPrior to Go Live CAISO will create an Operating Procedure that contains start-up values of parametersDuring first 12 months of operation, CAISO will issue updated versions of Operating Procedure as neededAfter 12 months CAISO will incorporate parameter values into a BPM and will utilize BPM Change Management to update values


Analysis Results for IFM Parameters

Presentation by Jim Price

Parameter Tuning for Uneconomic Adjustments - … · 2018-02-10 · Uneconomic Adjustment Scheduling Run and Pricing Run ... (Self Provision) – which include only MW quantities,

Documents