Provide your organization’s comments on the presentations provided by stakeholders at the working group:
Data & Analysis for the ESE Initiative
PG&E generally agrees with other stakeholders that the ESE initiative should be “data-driven” and analysis and modeling is needed to reinforce any market design enhancements. PG&E suggests that a partnership of CAISO, stakeholders, and one or more consultant organizations, partner with the CPUC and/or CEC (who are also stakeholders) to formulate a clear plan of analysis to be performed over the next year. Considering the total operational storage capacity in the CAISO has exceeded 1,000 MW, 2021-2022 data would prove invaluable to this effort.
In regards to stakeholders’ suggestion to publish advisory price data, PG&E agrees that this action may help build confidence in CAISO storage optimization. However, given the large volume of data involved, PG&E asks whether it would be preferable to publish advisory prices on an ad hoc basis (as has been done, to some extent, with mileage accuracy data for regulating resources) rather than creating standard publication reports. With respect to the impact advisory prices have on Bid Cost Recovery (BCR), PG&E addresses this topic in the BCR section of its comments.
Multi-Interval Optimization (MIO) & Market Signals
PG&E’s assumption is that storage’s energy and ancillary service values are maximized by participation in the day ahead market, and that marginal changes in real time, although they will add value, will not dominate battery bidding strategy. Gridwell/WPTF’s presentation made the assertion that “rational” real time market prices maximize the value of storage. PG&E disagrees with Gridwell/WPTF’s principle that prices along are enough to guarantee reliable system operations in a system with marginal cost signals not based on thermal heat rates.
In response to LS Power’s concerns about MIO, it might be possible to use SOC management in day-ahead optimization and then not in real-time. In effect, such a resource model would treat batteries as hybrids in real-time, but would use SOC management to establish robust cycling schedules based on the more predictable arbitrages in a day. Resources would have to manage to risk of no SOC management in the same way that hybrids do, through their bidding and possibly using a dynamic limits tool, but the “thirty minute buffer” CAISO requires for regulation awards might be sufficient for the purpose. This resource model would not allow for MSOC or target SOC management in real-time. Because dispatch would not involve cross-temporal effects (i.e., future forecast prices would not affect the current hour’s schedule) using this model, BCR and scheduling would be unaffected by MIO.
Another option, certainly not feasible within the current scope of market enhancements but possibly appropriate for addressing the issue, would be to run binding multi-interval optimizations for successive intervals of multiple hours. At one time the California electricity markets included a six-hour Day-Of Market, and there’s no reason that approach couldn’t be used to minimize uneconomic instruction due to the rolling MIO horizon. Under this option, there would be reduced impact of nonbinding future schedules (more along the lines of the day ahead market, where day 2 RUC can influence market results but only on the margins), and BCR would be calculated for the Day-Of horizon, with residual FMM and RTD instructions likewise having reduced impact on commitments and battery cycling.
Variable Costs, Real-time Bidding Enhancements, Spread Bidding
In their presentation, CESA asserts that current market bidding functionality does not allow batteries to precisely reflect cycling costs, even if additional cycling could be economic to meet reliability needs. PG&E doesn’t understand this assertion since battery cycling costs are fundamentally a calculation of long-term future opportunity costs versus an equally uncertain profile of long-term battery degradation. Therefore claiming that batteries know enough to precisely reflect cycling costs in their bidding is questionable. The variable cost component of the battery Default Energy Bid provides a way of agreeing on a single dollar value with CAISO, modifiable on an ad hoc basis if evidence is available that the component’s value has changed; this value allows a trade off in the market between additional cycling for reliability and battery preferred usage profiles.
CESA also requests for the CAISO to allow storage to submit multiple RT bid curves that are dependent on SOC/cycle. PG&E offers the following comments on this proposal:
- The CAISO cannot dynamically change bids based on SOC within the market processes, so changes in SOC as a result of optimization cannot cause changes in the bids used. If CESA is proposing that SOC ranges be treated as “configurations” like those of MSG resources, such a discretization would tend to unduly restrict flexibility in the real-time use of batteries.
- The CAISO could, however, choose one bid curve from a set of multiple RT bid curves, based on initial SOC or cycles in the day so far. This would be an increase in the size of the RT bid set to be validated, and would require clear bidding rules as to how bids for the “in-between” SOC levels (e.g. state of charge levels that are between multiple bids) would be determined. Additional rules would have to be establish regarding which default bids to be used in case SOC telemetry was lost or SOC was outside the range of the bid curves submitted.
- PG&E recommends a possible alternative to this approach: using some form of the Dynamic Limit tool that is adapted for SOC NGR. This method would allow for more frequent bid updating capabilities (in line with CESA’s request) and be more feasible in terms of implementation compared to the multiple bid curve approach.
In their presentation, GDS proposed an enhancement which would allow energy storage resources to update market bids as close to the fifteen-minute market execution as possible. PG&E has the following comments on this proposal:
- Real time bidding enhancements in the CAISO markets should allow for quantities that are offered to be modifiable, subject to constraints on economic withholding (Note: this assumes RA dollars and requirements will act as sufficient incentive to prevent such withholding). Bid prices, however, should not be modifiable unless all market participants have the same ability to modify their prices on a given timeline. If it’s possible to enforce this principle, PG&E sees no reason not to support the proposal by GDS “in principle.”
- It may not be possible to support this principle in cases where changing the upper or lower range of a bid effectively causes the market to clear on a different bid segment, as this may have the effect of allowing changes in bid price in cases where not everyone has the ability to modify their bid price.
- The enhanced timing of bid submission needs to be robust enough to handle large numbers of bids being processed by SIBR closer to the market process run, so this proposal should be based on a “worst case” in which all bids are resubmitted to SIBR on a nearer to market timeline. In other words, the time required to process all bids in SIBR, at least when there are no market-breaking business process rule violations, should set the minimum time before market that bids may be submitted.
- Depending on SIBR processing times, it may make sense to enable fifteen and five minute bids to be resubmitted into the CAISO markets on the same timeline as that of the first FMM run currently (something like twenty-seven minutes before market run, rather than the current seventy-five). However, one implication of such an enhancement that must be looked at closely is whether LMPM will then need to be re-run every fifteen minutes, rather than only once per hour.
GDS also proposed an enhancement to allow energy storage resources to provide a cost adder that is a function of average state-of-charge and depth-of-discharge over the market horizon. PG&E has the following comments on this proposal:
- If this is a proposal for an adder that changes dynamically within the optimization based on average state-of-charge and depth-of-discharge over the market horizon, such a proposal is not feasible using the current market optimization algorithm, because it introduces a nonlinear interaction between the adder and the battery schedule within the market algorithm.
- If this is a proposal for a real time adder based on day-ahead market results, this would arguably be a topic for discussion in enhancing the Default Energy Bid calculation.
- If this is a proposal for post-market cost recovery, this would arguably be a topic for discussion in enhancing Bid Cost Recovery for batteries; but it would also imply that battery bids inherently cannot account for such costs, even though other resource types are subject to similar costs that cannot be expressed exactly through bid curves but are able to participate without issue in the markets. Moreover, it would imply that the market solutions themselves would be suboptimal even prior to the additional cost recovery provided to batteries. Because batteries can adjust their bidding over time, PG&E doesn’t believe either of these implications to be true.
LS Power proposed in their presentation to make spread bidding optional for storage, given the MIO “promises” a spread but may not realize it in the market. The second motivation for the proposal is that despite common belief, a single spread bid doesn’t actually capture a spread for any pair of charge and discharge prices. PG&E agrees with this concern. Because charging costs must be divided by the roundtrip efficiency factor to get the effective charging cost per MWh of increased SOC, the arbitrage required at different charging cost levels differs. Additionally, the anticipated mitigation of discharge bids will override any spread component of market participant bids.
PG&E recommends that one solution to this concern is for the CAISO to offer an NGR model without SOC management for participants who don’t want their charge/discharge bids to have any resemblance to spread bids.
Ancillary Services, SOC Firming Product, Foldback Constraints
PG&E shares the concern that Gridwell/WPTF raised with respect to the definition of regulation range for hybrid resources. PG&E has long noticed this issue would hamper regulation and spinning reserves being provided by solar resources. PG&E is interested in the idea of a range not held to absolute maximum and minimum levels; however, such an approach requires a method of testing that would allow the CAISO to have confidence in a dynamic regulation range based on the level of solar production by the hybrid or solar resource. Perhaps a series of tests could be used to demonstrate such capability.
The concern that Gridwell/WPTF raised regarding infeasible regulation awards, however, is not shared by PG&E. First, CAISO has implemented a requirement that batteries have a thirty-minute energy buffer in order to provide regulation in the constrained direction, which should prevent most such infeasibilities from occurring except under unusual circumstances. Second, any risk that remains can be incorporated into bids just as regulation energy take can be priced into bids.
PG&E’s proposal for an Ancillary Service SOC Firming Product addresses many of stakeholders concerns regarding limited market horizons and the replacement of the MSOC. Opportunity costs of shifting energy from DA awarded hours to other hours (captured through price arbitrage for other energy-limited resources) cannot presently be compensated for SOC-managed resources. CAISO’s management of SOC imposes limits on dispatch awards beyond those contained in bids alone. PG&E’s described the following proposal for its SOC Firming Product in its presentation, but reiterates them here for emphasis:
- In the current market timeline, SOC firming can reasonably be procured only in the day ahead market.
- SOC firming requirements must thus be defined in advance of, and so independent of, day ahead market awards on SOC-managed resources.
- SOC firming imposes a minimum SOC on awarded resources. This minimum SOC must be consistent with market awards, and should be published along with market awards.
- For SOC-managed resources, an SOC firming award will be enforced in the real time market processes (using the existing MSOC logic). Enforcement may result in uneconomic charging, which may or may not be subject to BCR. PG&E’s straw proposal is that uneconomic charging should be not subject to BCR as long as SOC firming bids are compensated.
PG&E views an SOC Firming Product as preferable to an Energy Shift product (proposed by CESA) because the Energy Shift could be interpreted as a payment for maintaining a MSOC. Thus, batteries that receive “awards” have absolute market power to charge any price they like (in theory). It is therefore likely that there would be administrative limits on the product’s price, and those limits would effectively set the price, since there would be no benefit to a battery in offering the shift product for anything less. There would also be no need to impose an additional self-scheduling requirement if the MSOC logic is functioning correctly.
Gridwell/WPTF’s assessment is that use of the end-of-hour (EOH) state of charge parameter will potentially obviate the need for CAISO procurement of a storage product, or at least imposition of constraints similar to MSOC in the future. PG&E asserts that the horizons of decision-making by market participants and CAISO are always going to be different enough that market participants will not be able to simultaneously satisfy their own and CAISO’s objectives. Moreover, the EOH SOC parameter has a major “weakness” (actually it’s a feature) in that it disqualifies significant portions of the real-time operating horizon from BCR.
While PG&E is sympathetic to Vistra’s expressed need for CAISO to represent foldback limits in its modeling of batteries, it is doubtful that such modeling can be introduced without making significant changes to the basic SOC management model. This kind of effort would require a stakeholder process (along with other battery model change requests) as well as be vetted against computational tractability. Thus, PG&E does not support modeling changes to support foldback constraints in the time frame of the next year.
Bid Cost Recovery (BCR)
The topic of BCR is crucial to the ESE initiative since it will determine whether the unique aspects of energy storage resources warrant different rules for compensation compared to traditional resources, and if so, how those rules will be structured.
In their presentation, LS Power emphasized the point that nonbinding prices are being used to determine (uneconomic) binding awards. While the same issue affects all CAISO supply to some degree, the question the ESE initiative needs to address is: are SOC-managed NGRs significantly more impacted by this than traditional generators? Certainly, other generators do not have states of charge and/or receive charging awards. But before making a fundamental change to market design, (i.e. by removing or limiting multi-interval optimization (MIO) for storage) more data and analysis should be solicited by the CAISO.
A more feasible solution would be to guarantee BCR for binding instructions based on the advisory schedules used to generate the instructions. As an example, suppose a battery was charged at $30 based on a $40 bid based on an optimization that discharged the battery at $60 based on a $50 bid in a nonbinding interval. The recoverable cost associated with the instruction, assuming the discharge instruction was reversed when the optimization was run for its binding interval, would be $30 (the lost opportunity cost in the market algorithm) versus the current calculation of -$10, a reduction in BCR even though the charging was not used. A later binding discharge instruction would effectively reduce that BCR at any price above $60, and add to recoverable costs at any price below the $50 bid. This approach would require the following: (1) CAISO to publish all advisory schedules and prices associated with binding intervals, (2) a BCR calculation that would move through the window of real time optimization horizons and (3) significantly more complex calculations than current BCR calculations.
A similar but simpler option would be to enable real-time BCR to be calculated over successive battery cycles, defined as successive points at an identical state of charge with significant amounts of charge and discharge occurring over the interval. All batteries necessarily see such cycles in their daily operations, and they could be identified based on either a midpoint SOC or an initial SOC for a given day.
A third option would be to base BCR for NGRs on discharge revenues netted over the day, minus actual charging costs, minus any loss due to charging at above the charge bid level. In this calculation, there would be no offset specifically for actual charging costs being below the charge bid level, on the assumption that the benefits of this “profit” can only be realized when the battery discharges. Comparing this method to BCR for thermal generators, there are no adjustments made for generators who happen to have bought gas at below the CAISO’s index. This approach could also be used with exceptional dispatch BCR, based again on actual charging costs compared to default energy bid charging costs. (Note: the term “default energy bid charging costs” refer to the charging costs used in calculating the default energy bid (DEB), along with roundtrip efficiency and cycling or degradation costs.)
PG&E has several questions on CESA’s proposal to “Net all costs to charge the resource with the revenue from discharging to ensure bid spread recovery”:
- Does this proposal imply that real-time and day-ahead charging and discharging costs would be combined in the calculation of BCR?
- If a battery starts at SOC point A and ends the BCR calculation at point B, how are these points incorporated into the BCR calculation? Is a charging cost imputed to initial state of charge? If there’s a charging cost to move the battery to point B, is that cost recoverable even though it hasn’t yet been used for discharge?
- Would BCR potentially be calculated across days?
- How would self-schedule or target state of charge hours be removed from this calculation?
As several stakeholders mentioned during the Workshop, there are other CAISO initiatives which will likely overlap, or be combined, with the ESE initiative. Two examples are the ESDER and Hybrid Evolution initiatives. PG&E urges the CAISO to be adaptable in its policy development to handle such overlaps. Specific topics that came up at the workshop include:
- Long Duration energy storage—PG&E agrees with Gridwell/WPTF that long duration energy storage should be incorporated into the Energy Storage Enhancements (ESE) initiative on a 2-3 year timeline. Even four hour batteries take on attributes of long duration storage if there are enough of them in the system, in that there’s a need to account for arbitrages beyond the market model horizon, whether it be 24 hours or one of the real time market horizons.
- Batteries with discrete transition decisions and costs—PG&E agrees with Gridwell/WPTF that enhanced modeling of batteries with discrete transition decisions and costs (based on real physical operating characteristics, not strategic bidding considerations) will be valuable for incorporating post-lithium ion storage technologies into the markets. PG&E also suggests investigating whether the pumped storage model can be used as a basis for such modeling, as it already includes some features of this sort and is less constrained by state of charge modeling that may not be valid for some new storage technologies.
SDG&E’s hydrogen storage system appears to be similar in some ways to pumped storage, in that there is neither a need for the CAISO to manage the resource to a fixed maximum SOC (as this could easily change, with hydrogen being a reservoir of fuel energy constrained only the number of tanks available to store its production), nor a true continuous transition between discharging and charging. The hybrid model is certainly one approach to modeling the resource. It is important to note that fuel cells have not typically been treated as continuously dispatchable resources previously in the CAISO markets. This type of storage resource may point to a larger category of non-SOC managed resources that would include pumped storage and other technologies with no clearly defined storage reservoir, and possibly discrete state changes with associated costs.
Provide any additional comments on the working group, or any additional scope items your organization feels should be included for this initiative. You may upload examples and data using the “attachments” field below: