Stakeholder Comments Pay for Performance Regulati ion Straw Proposal Submitted by Company Date Submitted Kai Stephan (626) 302-3571 Alex Morris (626) 664-9926 Southern California Edison (SCE) December 19, 2011 SCE appreciates the opportunity to provide comments onn the Pay forr Performance Regulation straw proposal. SCE supports the general direction of the Californiaa Independent System Operator s (CAISO) proposal but believes this design effort should include a more comprehensive review of Regulation, its integration withh Flexible Ramping Products, and its intended role in grid reliability (addressing intra-rtd errors or system trending) ). In this design process, the CAISO s goal should be to design Regulation correctly. This goal should not be sacrificed in meeting FERC s timeframe. Elements off the ultimate design may need incremental implementationn to comply with FERC deadlines. I. SCE urges the CAISO to discuss critical elements not in the proposal. A. CAISO needs to disclose its Regulation dispatch algorithm to enable effective stakeholder input. Stakeholders need to better understand the CAISO s dispatch algorithm and how it operates with energy schedules in RTPD and RTD and with Flexible ramping dispatches. In the CAISO stakeholder conferencee call, the CAISO stated the objective function seek(s) to maximize future available ramp. 1 In order to provide effective input to help design an efficient Regulation product, the CAISO needs to disclose to stakeholders their Regulation dispatch algorithm. SCE also suggests the CAISO hold a technical workshop on the dispatch algorithm. B. Regulation should solve for intra-rtd uncertainty only, while flexible ramping services and imbalance energy markets resolvee trending, variability, and sub-rtpd uncertainty. Data providedd during the Regulation Energy Management (REM) stakeholder process indicates Regulation serves to meet system trending needs. In light of the developing roles for flexible ramping services, SCE recommends Regulation be designed specifically to resolve system noise the smaller random deviations resulting from intra-real-time 1 CAISO Stakeholder conference call on Frequency Regulation, 1:00pm December 12, 2011 1
Dispatch uncertainties. Regulation should no longer be used for system trending. Instead, the Regulation signal should resolve small up and down fluctuations resulting in a zero-net-energy (ZNE) type of product. Separately, system trending and intra-rtpd uncertainties should be solved using the combination of Flexi-ramp and Real-time dispatch schedules. With a ZNE Regulation product, any residual energy would be settled at the resources default energy bid. This design immunizes resources from Real- Time energy price exposure. A ZNE product would likely require a single Regulation product design. SCE discusses this concept later in these comments. The CAISO should minimize the capacity procured for system variability. If the CAISO can manage longer term variability and uncertainty with products typically less expensive than Regulation, it should do so and end procurement of regulation capacity for such functions. Data and experience gained from the flexible ramping services should inform the CAISO s regulation design as a noise product. C. An efficient design requires costs are allocated properly. FERC set an important precedent by approving Westar s Portfolio s cost allocation methodology. CAISO should consider this precedent in forming its cost allocation proposals. Costs for integrating and balancing services should go to those causing the need. SCE recommends the CAISO calculate Regulation needs caused by each market participant (i.e. Load, VERs), and allocate procurement costs accordingly. Since Regulation and ramping procurement will increase through 2020 as the CAISO integrates VERs commensurate with the 33% Renewable Portfolio Standard goal 2, the CAISO should develop proper cost allocation structures now so goals are achieved most efficiently and at the least cost to ratepayers. SCE recommends the CAISO adopt the FERC approved cost allocation methodology developed by Westar 3. Westar s portfolio methodology logically allocates the costs of balancing products to those causing the need using a statistical analysis of historical performance data. A presentation discussing the Westar s Portfolio methodology is attached to these comments. 2 CAISO Summary of Preliminary Results of 33% Renewable Integration Study ; May 10,2011 3 Westar s filing with FERC: ER09-1273 Page 2
II. SCE s comments on the specific elements in the current Proposal. A. SCE supports multiple aspects of the current Proposal. Specifically, SCE supports the following: Calculating mileage as the absolute change in AGC set points between four-second intervals. Payments for actual mileage moved, rather than for an expected mileage movement. No-pay for non-performance to the CAISO s dispatch signal. Mileage calculations that incorporate four-second interval instructions. B. The current no-pay calculations may be overly fine and discrete. Any performance metric should primarily measure performance and not measure error including communications lag. In designing accuracy and no-pay provisions, the CAISO needs to discuss how Regulation is actually solving area-control error, and ensure its design aligns with actual grid operations. Four-second performance and response measurement may be too granular because of response times or other normal aspects of Regulation dispatch and response. If a resource is inaccurate to every four-second dispatch signal, yet on 30- second horizon the resource is accurate to the cumulative instructions, has the grid received comparable benefit to four-second accuracy? If yes, should the resource receive a comparable mileage payment? Understanding CAISO s Regulation dispatch algorithm will help stakeholders answer these questions. The CAISO should not design a system where the majority of resources will incur no-pay provisions for every interval. No-pay for non-performance should primarily result from actual non-performance, and not from an accumulation of system errors. This structure ensures that resources are properly compensated for their mileage provided. Additionally, the change to mileage payments should leverage existing systems and not trigger costly new communications infrastructure expenses. SCE believes the CAISO s current proposal could trigger new expenses. As a principle, resources should not be penalized for infeasible dispatch instructions. For example, if it takes more than four seconds to receive the dispatch signal from the CAISO, generators can never respond accurately. Any No-Pay from inherent system and communications lags within the AGC system should be minimal. To this end, some form of deadband accuracy adjustment appears necessary. Without such a structure, system communication lags may be misinterpreted as non-performance. SCE requests CAISO provide additional information on timing and communication lags in the AGC system before deciding upon the proper settlement interval. Page 3
In the CAISO s current proposal, the mileage provided by a resource s governor response system may be considered No-Pay. Mileage produced from governor response system is usually helpful to the CAISO in correcting ACE, and therefore should be included in a mileage payment. 4 C. CAISO should explore a procurement design that explicitly constrains mileage and capacity. The proposal assumes sufficient mileage capability will automatically result from the procured Regulation capacity, yet this outcome is not guaranteed. The CAISO needs to convince stakeholders that this approach will sufficiently meet mileage needs. The CAISO should consider a procurement target and constraint for mileage capability in its procurement objective function. This approach eliminates uncertainty about the mileage that results from procurement solely for capacity targets. Without a mileage constraint, the CAISO will be unable to ensure it has sufficient mileage capability. Additionally, several potential problems may result from the proposed Mileage Multiplier. The average mileage capability from a previous day s resource mix may not represent the mileage capability of the current mix. Without knowing how much mileage each resource can provide, and without having a mileage procurement target, the CAISO may procure insufficient mileage 5. SCE notes alternatives to the Mileage Multiplier later in these comments. D. A proper Regulation design should reduce costs. The CAISO should seek a Regulation design where procurement of faster ramping resources reduces regulation costs. FERC cited the ability of faster-ramping frequency regulation providers to lower costs to load. 6 This savings should result in part through ISO/RTO procurement of less regulation. 7 The Commission reiterated this in its Final 4 A governor will read the frequency at its specific resource, and automatically adjust the resource s generation such that it stays as close to 60Hz as possible. Given that this system reads the frequency at the resource itself, it often will begin to adjust the resource s generation to meet 60Hz before the CAISO s Regulation dispatch signal has been received. 5 Day 1 requires 100MW capacity for Regulation. The CAISO commits two 50MW resources that each have a ramp rate of 5MW/min, and a total maximum daily mileage amount of 14,400MW-4sec. Day 2 also requires 100MW capacity for regulation. Using the previous day s Mileage Multiplier, the CAISO believes it can get 14,400MW- 4sec worth of mileage. However, the CAISO now commits one 100MW resource with an 8MW/min ramp rate, and thus a maximum daily mileage amount of 11,520 MW-4sec. Although the CAISO has met its Regulation capacity requirement, it is now deficient in mileage capability. 6 In the Frequency Regulation Compensation in the Organized Wholesale Power Markets, Notice of Proposed Rulemaking (NOPR), February 17, 2011, 134 FERC 61,124, RM11-7, Page 20, Paragraph 32 7 Demonstrated by the Pacific Northwest National Laboratory study, also cited in the NOPR, Makarov, Y.V., Ma, J., Lu, S., and T.B. Nguyen, Assessing the Value of Regulation s Based on Their Time Response Characteristics, Pacific Northwest National Laboratory, PNNL-17632, June 2008 Page 4
Rule. 8 To SCE, an increase in regulation costs could imply an inefficient methodology or design. E. If the Mileage Multiplier calculation is pursued, it will require additional design. If the CAISO continues to use Regulation for system trending, the Mileage Multiplier should be calculated at least hourly rather than as a daily average. Calculating hourly would reduce the likelihood of procuring insufficient mileage for the entire day due to averaging out of the mileage need. Since capacity is the driver in the optimization process, mileage may be insufficient with a day-long mileage multiplier. Additionally, the Mileage Multiplier should be calculated separately for each class of resources (e.g. hydro, peaker, etc.), enabling more accurate representation in the procurement objective function. III. Rather than the proposed incremental changes to Regulation, the CAISO should consider a comprehensive redesign. SCE recommends the following ideas for consideration in this process. Alternative approaches on Pay for Performance Regulation design exist. SCE does not necessarily advocate for these alternatives but believes them worthy of stakeholder discussion. An attached presentation focuses on these concepts. A. CAISO should explore Regulation as a single product. As is done by all other ISO/RTO s, Regulation Up and Regulation Down can be combined into a single Regulation product. The current split product is somewhat artificial once a resource provides Regulation Up, it essentially provides Regulation Down when dispatched back to its original set point. s that only provide Regulation in a single direction could have their initial set-point biased to their upper or lower range, allowing provision of both up and down Regulation. A single Regulation product used in this way greatly reduces a resource s exposure to Real-time energy prices, reducing risk premiums. In this structure, all Regulation energy production/consumption should net to zero over time assuming the dispatch algorithm is designed accordingly. A resource s default energy bid could be used to pay/credit resources for net production/consumption. 8 In the Frequency Regulation Compensation in the Organized Wholesale Power Markets, Final Rule, October 20, 2011, 137 FERC 61,064, RM11-7, Page 12, Paragraph 19 Page 5
B. CAISO should discuss a turnaround rate as a measure of movement capabilities of resources. With a Regulation product designed to resolve noise, regulation needs and corresponding dispatch signals may move up and down multiple times within a 5-min interval. for such movements benefits from the use of a turnaround rate, defined as the maximum movement (up or down) a resource can achieve while still returning to its starting point within a five minute interval. The number of turns possible within the 5 minute interval would also be measured (i.e., a battery may be able to reach its max MW output and return to zero several times within 5 minutes). These turnaround metrics, in addition to a general capacity measure, better reflect a resource s mileage capability, now defined as its turnaround level multiplied by its potential number of turns. s would report their physical max turnaround level (in MW), and the number of times they can achieve that level within 5 minutes. The turnaround rate is preferable to the current Mileage Multiplier because it better reflects a resource s ability to move up and down. Therefore, the turnaround rate should be used to define a resource s mileage capability in the procurement objective function. C. Once constrained for mileage/turnaround, the objective function costs should be multiplied by an expected-use quotient that is resource class specific and calculated from historical data. Prior to real-time, the actual mileage a specific resource will provide is unknown. Thus, the procurement objective function needs to estimate a resource s mileage in order to minimize costs. To best estimate a resource s mileage, an expected-use quotient can be used based on that resource s class (i.e. hydro, battery, flywheel, CCGT, etc.). This quotient can be calculated by resource class from the rolling average of historical mileage production. This approach is preferable to a Mileage Multiplier because mileage expectations are more logically estimated and represented in the system optimization. Page 6
Summary comparison of CAISO s proposal and alternative designs. Feature CAISO Alternate Proposal Regulation Procurement Expected use of Reg product Real-Time Energy Cost Allocation Procurement constraints Defining resource s mileage capability Separate Reg Up and Reg Down products System trending and system noise Reg resources exposed to RT energy prices Load pays everything Only have a capacity targetassumes sufficient mileage will be provided Averages a previous day s mileage provided by one resource pool, assumes all resources can provide exactly previous day s average mileage Single Reg product System noise only- Zero Net Energy product Little excess energy due to ZNE dispatch algorithm, remaining excess settled at the resources default energy bid Costs allocated based on causation, per Westar s Portfolio methodology 9 Explicitly constrains procurement minimum on mileage and capacity requirements Specifies unique mileage capability for each resource using a turnaround rate- a physical characteristic of each resource 9 Westar s filing with FERC: ER09-1273 Page 7
APPENDIX Page 8
Westar s Portfolio Cost Allocation Methodology December 19, 2011 Jeffrey Nelson (626)302-4834 Jeff.nelson@sce.com
Agenda Portfolio effect and basic stats review Westar- Portfolio Approach Presentation Objectives Gain understanding of various cost allocation methodologies Draw conclusions on what might work best for CAISO market Page: 1
Statistics Review Standard Deviation, σ σ 2σ 3σ 68.2% of values 95.4% of values 99.7% of values X Y X Y X Y Correlation, ρ High positive correlation, ρ 1 High negative correlation, ρ 1 No correlation, ρ 0 Y Y Y X X X Covariance, Page: 2
Big Picture General Problem: The total system regulation requirement is known, how do we allocate this total based on individual resource/ load contribution to need? Portfolio Benefits: System total sum of needs to manage each individual resource/ load Portfolio almost always less (rarely equal, never greater) This reduction is commonly called Portfolio Affect, Diversification, or Offsetting Errors Page: 3
Example: Portfolio of Wind and Load Load Wind Probability Probability 2σ 300 MW 2σ 110 MW MW Assume BAA is carrying enough Regulation reserves to cover 2σ of uninstructed movement (95% of all events) Without benefits of a portfolio, typical BAA would have to carry: However, due to offsetting errors, the BAA s portfolio needs less regulation reserves MW Example: 2 = 300 MW 2 = 110 MW = 410 MW Page: 4
Example: Portfolio of Wind and Load (cont.) To calculate adjusted variance, need standard deviation (σ) and covariance between resources Load Wind Adjusted Variance Load,, Wind,, The Portfolio s standard deviation is calculated as the sum of the adjusted variances:,, Example (cont.) Load Wind Adj Var Load 22,500 1,650 24,150 Wind 1,650 3,025 4,675 Assumed 0.2 24,150 4,675 Procurement = 2 340 MW The Portfolio s standard deviation is always less than or equal to the sum of the standard deviations of its individual components 1 : Note: Portfolio procurement quantity reduced to 340 MW from Standalone approach s quantity of 410 MW 1- See Backup slides for proof Page: 5
Example: Portfolio of Wind and Load (cont.) Each individual resource/ load is then allocated its share of the portfolio s regulation needs Pro-rata Formula (e.g. Wind s Share): Wind s Allocation = Page: 6
Example: Portfolio of Wind and Load (cont.) Each individual s share is a pro-rata based on adjusted variance: Load s Share: 2 Wind s Share: 2 Example (cont.) Total Regulation Procurement = 2 340 MW Load s Share = MW Compared to 300 MW in Standalone approach Wind s Share = MW Compared to 110 MW in Standalone approach Page: 7
Westar- Portfolio Approach Uses correlation between individual resources/ load Same calculations as example on slides 4-6, but with 5 components instead of 2 Calculates total portfolio as: 2 System Diversity Ratio Westar s Results (as presented to FERC): Westar redacted the data in their public report 1 Share to wind drops to 7.25% Recall: share to wind from standalone approach was 7.8% 1 Puget Sound Energy adopted Westar s Portfolio methodology, and provides more data for their calculations: http://elibrary.ferc.gov/idmws/common/opennat.asp?fileid=12674122 Page: 8
Westar Observations Only thing needed is historically deviation data CAISO already has 10-min generation data, 1-hr load data Load data can be profiled Very robust and easy methodology Allows for an unlimited number of buckets (can break down allocations to specific units) Calculated on an hourly basis Need to account for all instructed deviations i.e. instructed deviations should not be counted for cost allocation CAISO has data and can easily implement Page: 9
SCE: Current thinking on Frequency Regulation December 19, 2011 Jeffrey Nelson (626)302-4834 Jeff.nelson@sce.com
Overview Single Regulation product (not separate Reg Up and Reg Down) Regulation is used for noise not trending, thus it is expected to be Zero Net Energy (ZNE) CAISO determines the mileage capability of each unit as part of the selection process CAISO specifies both capacity needs and mileage needs Page: 1
Regulation Market Design Framework Regulation Dispatch Signal to continually cross zero Regulation should capture noise, and not be used for system trends Separate following product should be developed for broader trends Single Regulation Product: Combine Reg Up Reg Dn Reg energy production/consumption should net to zero Eliminates/ greatly reduces exposure to RT energy prices Procurement for Regulation Procure enough capacity for expected Reg magnitude - based on historical statistics Procure enough mileage for how much Regulation is expected to move- based on historical statistics Settlement for Regulation Paid for any non-net zero energy based on resource production cost/ default energy bid (not market) Pay for capacity, via (Capacity Market Clearing Price)*(Capacity Sold) Pay mileage, via (Movement Market Clearing Price)*(Movement Sold or Produced) Movement must comply with ISO dispatch signal (pay for performance) Operational Characteristics represents:1) total 10-min ramp capability [can substitute for other ancillary services], 2) 5-min Turnaround rate, and 3) number of turns in 5-min Based on true physical characteristics Bids Price to reserve capacity Price per MW moved ( mileage rate) Page: 2
Regulation Visual 350 300 Real-time Ramping Energy Actual Load During this interval, Regulation is needed to peak Up AND return Down to near original set-point MW 250 200 150 Regulation is used to fill the gaps between Real-time Ramping Energy and Actual Load 0 1 2 3 4 5 6 7 8 9 10 11 12 Interval Page: 3
The Turnaround Constraint MW 350 300 250 Real-time Ramping Energy Actual Load Reg Up Reg Dn 200 5-min 150 0 1 2 3 4 5 6 7 8 9 10 11 12 Interval CAISO currently does a ramping constraint calculated by how fast a resource could ramp up in 10 minutes In reality, a proper Regulation dispatch signal 1 will move up AND down during a 5-min interval Note: CAISO dispatches energy in 5-min blocks Use a Turnaround Rate rather than a Ramp Rate for mileage estimate i.e.: Starting at a set point, what MW max can a resource meet- and still return to the set-point- within 5 minutes 1- The Regulation dispatch signal should be designed to compensate for the system noise, whereas Flexi-Ramp should be used for all system trending/ ramping Page: 4
Turnaround Constraint Examples 60 50 40 Conventional Ramp Up Turnaround 60 MW Here, the conventional resource can reach 60MW output in 5-min Capacity = 60 MW MW 30 20 10 0 0 1 2 3 4 5 Minute 11 MW If that same resource needs to return to zero, its turnaround is the max output reached Turnaround = 11MW, 1 Turn MW 60 50 40 30 20 10 0 Battery Ramp Up Turnaround 0 1 2 3 Minute 4 5 20 MW The battery reaches its full output in only 30-sec, but is limited by its capacity Capacity = 20 MW Using the turnaround constraint, it can reach its max output several times Turnaround = 20MW, 5 Turns Page: 5
Procurement/ Bidding Structure s would state their capacity, Turnaround, and number of Turnarounds s would bid their capacity price and movement or mileage rate Example: Physically Based Market Based Capacity Turnaround Number (#) of Turnarounds Capacity bid [$/MW] Speed bid [$/#MW] A 60 MW 11 MW 1 $0.75 $0.30 B 10 MW 10 MW 5 $1.20 $0.10 C 5 MW 5 MW 25 $1.00 $0.02 Page: 6
Mileage Performance Definition Mileage is the absolute value of instructed change in operating point No mileage payment 2 miles payment 102 101 3 miles payment MW 100 1 mile payment 99 0 1 2 3 4 5 6 7 8 9 10 Minute Total mileage payment = 2 + 3 + 1 = 6 MW * Rate Page: 7
Zero Net Energy Design Dispatch Algorithm designed to result in zero net energy for a unit over some time horizon 102 101 Positive Energy MW 100 99 Negative Energy 0 1 2 3 4 5 6 7 8 9 10 Minute Set Point Positive and negative energy should net as close to zero as possible Remaining energy will be paid at resource s cost and charged to load Page: 8
Procurement Constraints ISO has Two Constraints: Total Capacity and Total 5-min Mileage Capability 1. Total 10-min uni-directional capacity e.g. 1100MW Reg Up, 1300MW Reg Down 10-min synchronizes well with other ancillary services 2. Total 5-min mileage capability 5 min e.g. total mileage CAISO needs over 5-min is: 15MW/min 10 ; 5 ; 15 /min If ISO purchases from a resource i, mileage calculated as: where,, Page: 9
Procurement Objective Function Sum of Capacity Payments Expected use See next slide Sum of Mileage Payments,,, Subject to: 1 2 Total mileage ISO finds a least-cost solution given constraints to determine marginal resources Shadow price of constraint = Capacity MCP Shadow price of constraint = Mileage MCP 1 2 Page: 10
Adjustment to Mileage Objective Function At time of resource selection, the actual mileage a resource will provide is unknown,, Adjustment factor for actual mileage Objective function cost should be adjusted based on expected use U i Either: 1. Set 1 for all resources Recognizes resources full capability won t be used on average 2. Define specific to each resource i or resource class i Use historic data to weight based on observations s are paid for their actual movement U i Page: 11
Procurement Example Bids Capacity Turnaround Number (#) of Turnarounds Capacity bid [$/MW] Mileage bid [$/#MW] A 60 MW 11 MW 1 $0.75 $0.30 B 10 MW 10 MW 5 $1.20 $0.10 C 5 MW 5 MW 25 $1.00 $0.02 CAISO Requirements: Capacity: 60MW Mileage: 50MW Capacity Awarded Capacity Cost Mileage Gained Mileage Cost A 58.439 MW $43.83 11 MW $3.30 B - - - - C 1.560 MW $2.34 39 MW $0.78 Sum 59.999 MW $46.17 50 MW $4.08 Page: 12