In the April 17, 2024, meeting of the Resource Adequacy Subcommittee (RASC), MISO discussed LOLE model improvements focused on planned outages (planned maintenance). Feedback is requested on the appropriate distribution of planned outages across all resource classes during the Critical Hours.
MISO requests that stakeholders provide supporting rationale and example(s), if applicable.
Comments are due by May 8.
As Duke Energy Indiana has expressed previously, we believe that including planned outages in the LOLE modeling while simultaneously maintaining the 31-Day Rule with capacity replacement requirements is double-dipping. That said, a middle-ground could help close the gap on this issue. To the extent the 31-Day Rule requires capacity replacement only for planned outage days in excess of 31, it may be reasonable to maintain the 31-Day Rule, if all planned outages as modeled in the LOLE were limited to 31 days or less. Thus, a planned outage representation would continue to be included in the LOLE model, but there would be no double-dipping as any replacement requirements would then be assessed on planned outage time that would not have been included in the LOLE model. Also, to the extent only such shorter periods of planned outage would be modeled, it would help to reduce the effect of those fixed periods on the modeling results, as discussed during the RASC presentation. Further, though, the 31-Day Rule is not specific as to which 31 days of a greater-than 31-day outage are “free” (i.e., do not require replacement). In as much as it could be the first 31 days, the last 31 days, or any combination of 31 discontinuous days within the entire outage event, the 31 days of such outage that could be included in the LOLE modeling could become random draws, just like forced outages. So planned outages of less than 32 days could still be time-fixed within the LOLE model, but planned outages greater than 31 days could be set to randomly draw 31 days from within the outage window. That concept could even further improve the probabilistic nature of the model, substantially resolve the issue that was discussed at the RASC, and at the same time cure the current double-dipping issue with the 31-Day Rule.
WEC Energy Group shares the concerns raised by MISO staff and stakeholders on the use of a maintenance schedule within the LOLE study that is highly correlated across the 30 weather years and across each Monte Carlo random draw of forced outages. An identical maintenance schedule across a large number of Critical Hours could inappropriately skew the Resource Class-level UCAP. We believe that the Maitenance Margin and the need to schedule outages more than 120 days in advance to recieve an exemption will drive mainteance outage behavior and that behavior will avoid planned outages during actual Critical Hours. A mechanism is needed to represent this type of planned outage behavior within the LOLE study for each weather year.
The Environmental Sector appreciates the opportunity to comment on MISO’s consideration of planned outage modeling within the LOLE model and supports MISO’s effort to take a deeper look at planned maintenance and its impact on accreditation. Yet, we believe that MISO’s question about whether the distribution of planned outages across all resource classes during critical hours should be the same as the distribution of planned outages across all resource classes across the season or year is premature and cannot be satisfactorily answered without more data. Stakeholders need more information to assess the impact of different approaches to distributing planned outages during the season or year. We offer the following comments/suggestions:
First, we heard MISO say during the April RASC meeting that the manner in which planned outages are represented is not material to LOL outcomes, but we’re not clear why that would be or what information MISO has evaluated to reach that conclusion. We request that MISO clarify why it believes this is true and that it share any related analysis or information with stakeholders. The January 2021 RASC meeting[1] included a discussion of possible outage modeling approaches including outages scheduled with perfect foresight of load, outages scheduled according to average load, and outages calibrated to historic schedules. This along with other discussions of planned outage scheduling in the LOLE model have led the Environmental Sector to the conclusion that planned outages are very important for model outcomes in the LOLE model and so we’re uncertain how/if to give feedback when it’s not clear how to judge the urgency and scale of issues caused by the manner in which planned outages are modeled.
Second, it's our understanding that planned maintenance is currently modeled as described in the Astrape memo[2] on supplemental season inputs dated July 8, 2022. Please provide any corrections if the below description does not match with MISO’s current approach.
Each generator is given a planned maintenance rate (maint_rate) in SERVM which represents a percentage of the year the unit is on planned or forced maintenance. The hybrid approach developed as part of RAN forces 80% of all the planned outages to be optimized based on analyzing the 30-year net load shapes but not having perfect knowledge of that year’s specific net load shape. SERVM takes the peak of all 30 net load shapes to determine the net load shape to optimize the first 80% of planned and forced maintenance. For example, on Jan 1, the net load peak of all 30 years is used, then it looks at Jan 2 and pulls the highest net peak across all 30 years and so on to create a full 365 day profile. The remaining 20% of all planned and forced maintenance outages are optimized based on that year’s specific net load shape giving SERVM perfect knowledge for the last 20% of planned and forced maintenance outages. This allows for a reasonable amount of flexibility in planned and forced maintenance to be utilized to better represent real time operations.
Third, it would be very helpful to see meaningful analysis of alternative planned maintenance schedules within the LOLE model, such as 80% of planned maintenance based on schedules for the next planning year that have been shared by the market participants and 20% using the hybrid approach described above. We recognize that when a stakeholder suggested that MISO should use the planned maintenance schedules submitted to MISO, MISO rejected that suggestion because it said it doesn’t align with the weather years. Please explain why that matters. Is it MISO’s belief that market participants are scheduling maintenance based on their knowledge of the weather year data in MISO’s LOLE model? What connection is MISO drawing between the two?
Lastly, the Environmental Sector, and we believe the stakeholders generally, do not have enough information on which to base answers to MISO’s question. We recommend that MISO first solicit ideas from stakeholders regarding different approaches to modeling planned maintenance schedules, with a particular view to minimizing reliability impacts, as calculated by either/both of tight margin hours and LOLE hours. MISO should also demonstrate the impact, if any, that the current planned maintenance scheduling has on reliability (LOLE and/or tight margin hours).
In conducting more analysis and providing the results to stakeholders, MISO should
We also urge MISO to consider that any conclusions regarding the above will likely depend heavily on the resource mix, and on the current state of energy sector evolution to a more renewable resource mix.
Respectfully submitted on behalf of the Environmental Sector,
Natalie McIntire, Sustainable FERC Project
[1]https://cdn.misoenergy.org/20210106%20RASC%20Item%2003c%20LOLE%20Enhancements%20(Outage%20Modeling)508758.pdf
[2]https://cdn.misoenergy.org/20220707%20LOLEWG%20Supplemental%20MISO%20Seasonal%20Inputs%20Documentation%20Astrape625466.pdf
WPPI supports MISO’s plan to perform additional investigative analysis on the magnitude of impact caused by current modeling of planned outages. We acknowledge that concerns may arise when the maintenance profile is identical across numerous critical hours, potentially overinflating the risk for resource classes that are on maintenance during said hours. While performing additional analysis, we urge MISO to compare the hypothetical distribution of planned outages in the model against the actual planned outages (to the extent they are available). Comparing model vs. actual distributions will help ensure each resource class has a planned outage distribution that reflects the actual pattern of planned outages that’s been seen.
Further, WPPI suggests MISO account for separate resource class distributions in their analysis. Differing outage characteristics by class won’t necessarily be reflected by a single outage distribution across all resource classes. For example, planned nuclear outages can be scheduled well in advance for a longer period, while classes like coal or gas may have shorter outages but can be more flexible on timing. These differences are material and should be treated as such. Accounting for each resource class’s planned outage distribution will only serve to increase the accuracy and reliability of the model.
This Feedback is submitted on behalf of the Louisiana Public Service Commission and the Mississippi Public Service Commission
At the April 17, 2024 RASC meeting, MISO recognized that changes are needed to the SERVM model because that model does not equitably distribute planned outages across resource classes during critical hours. It committed to investigate the impacts of the modeling deficiencies and develop proposed solutions. The SERVM model determines the total amount of planned outages in a day with the objective of minimizing impacts on system reliability. SERVM only considers the total number of megawatts in planned outage each day and does not attempt to allocate an even proportion of the total planned megawatts in outage to each Resource Class. However, not all hours are equal in the accreditation calculation. Randomly scheduling certain resource types during Critical Hours, which are weighted at 80%, likely will result in an inequitable allocation by scheduling some resources during critical hours and other resources on days without extreme weather. Different resource types in fact are equally likely to plan outages during critical hours, which are unknowable in advance.
MISO seeks stakeholder feedback on whether the distribution of planned outages across all resource classes during Critical Hours should be same as the distribution of planned outages across all resource classes during the rest of the season, the rest of the year, or some other. The Louisiana and Mississippi Commissions do not have sufficient information to determine at this time whether any of those choices are superior to the others. The Louisiana and Mississippi Commissions recommend that MISO perform the necessary analyses to determine the impact of the inequities in the planned outage modeling, to examine and compare the seasonal and yearly approaches ( and any other reasonable choices) to the current modeling and utilize the stakeholder process to discuss those results and develop a reasonable solution.
to: | MISO Resource adequacy Subcommittee |
from: | The Entergy Operating Companies |
subject: | LOLE Modeling Enhancement: Planned Outage Modeling |
date: | May 8, 2024 |
The Entergy Operating Companies ("EOCs")[1] appreciate the opportunity to provide feedback on Planned Outage Modeling in MISO’s LOLE study. The EOCs appreciate MISO initiating the stakeholder feedback process for this issue and believe MISO’s initial presentation lays out the issue well.
The EOCs believe that planned outage modeling assumptions in LOLE analyses should not result in such outage scheduling assumptions becoming a primary driver of class-level capacity accreditation. Planned outages are scheduled and managed so as to minimize reliability impacts. Moreover, individual resources already suffer accreditation penalties for outages scheduled without tier 1 and/or tier 2 exemptions. Further, outages longer than 31 days in a season already result in penalty costs.
Regarding MISO’s question posed to stakeholders on slide 11 of its April 17th RASC presentation, the EOCs recommend that MISO utilize annual planned outage rate inputs to distribute planned outages across resource classes during critical hours, provided that the model allocates outages equitably across resource classes during these critical hours. For example, a disproportionately high (relative to annual planned outage rates) fraction of resource class A outages scheduled over 2/17/2021 (slide 9 example) may harm class-level accreditation for resource class A unjustly. Seasonal planned outage rate inputs may increase the likelihood that this occurs since there may be seasonal differences in the historical distribution of planned outages across different resource classes. For resource classes comprised of a small number of units such as nuclear, the distribution of historical planned outages across seasons is driven by a small number of outages and may contribute to volatility in class-level accreditation year over year. MISO should also verify that the planned outage modeling methodology does not create a bias against resource classes with larger units, such as nuclear.
The EOCs further believe that MISO’s 80% fixed and 20% optimized outage profile methodology described on slide 5 is a useful feature that balances the reality of imperfect information when scheduling planned outages with the flexibility of moving outages in the operational timeframe.
The EOCs encourage MISO to investigate whether derating units on the input side may be an acceptable way to achieve equity across resource classes. Not only would this approach capture an annual average reduction in availability but would also avoid the added penalty to accreditation that might result from an unsuitably assumed outage in the LOLE analyses. If not, the EOCs encourage MISO to continue to work with Astrapé to best utilize SERVM features to ensure equity across resource classes.
[1] The Entergy Operating Companies are Entergy Arkansas, LLC, Entergy Louisiana, LLC, Entergy Mississippi, LLC, Entergy New Orleans, LLC, and Entergy Texas, Inc.
NEER Response to MISO Feedback Request on Planned Outage Modeling LOLE
Due: May 8, 2024
NextEra appreciates the opportunity to provide feedback in response to MISO’s request for the appropriate distribution of planned outages across all resource classes during Critical Hours. As a general matter, NextEra supports MISO’s efforts and further discussions around appropriate Planned Outage modeling. NextEra supports using an annual distribution value for Critical Hour planned outages.
The illustration below highlights maintenance across the year by resource class. (see attachment for illustration)
NextEra, as a signatory to the Comments and Protest of the Clean Energy Parties in FERC Docket No. ER24-1638, identified the need for the distribution of planned outages for all resource classes during Critical Hours to represent the distribution of outages across the entire year. MISO’s current use of maintenance and planned outage rates for each resource class in the LOLE model, except for nuclear units that are subject to fixed outage schedules, does not attempt to allocate outages to resources classes that are on outage each day. As a result, random scheduling decisions made by the model can significantly penalize one technology class while benefiting another, regardless of how that class performs. Moreover, because nuclear units are subject to a fixed outage schedule, yet other thermal facilities have an outage rate applied, nuclear units are affected disproportionately and unfairly.
Currently, the model only considers the total volume of MWs on planned outage each day. The model does not allocate an even amount of each resource class on outage each day. As a result, the randomized scheduling decisions made within DLOL can severely impact resource accreditation for one resource class.
For example, if the DLOL model disproportionately assigns planned outages to one resource class during extreme cold weather days – when loss of load hours are highest – that resource class will be assigned a lower accreditation value. And unjustifiably, resource classes whose planned outages were randomly modeled as scheduled on a different day will have a higher accreditation value. This is an unreasonable outcome, fails to align planned outage scheduling with reasonably expected operations, and uses a randomized outage scheduling practice to separate reliability value from accreditation.
As an example, the resource class Planned Outage are represented in the Table below for the spring season. There is a meaningful disconnection when comparing each resource class’s total share of maintenance to that same share of maintenance during Critical Hours.
Resource Class | Share of Total Maintenance | |
Spring Average | Spring Critical Hours | |
Coal | 22% | 16% |
Gas | 35% | 22% |
CC | 29% | 57% |
Nuclear | 13% | 6% |
To resolve this disconnection, an annual maintenance outage rate should be adopted.
NextEra looks forward to working with MISO staff to further define the process for allocating planned outages across all resource classes during Critical Hours. To further this important discussion, NextEra requests that MISO provide additional analysis of Resource Class DLOL values under (1) seasonal distribution of planned outages in Critical Hours; and (2) annual distribution of planned outages in Critical Hours. NextEra requests that MISO also provide a roadmap of ongoing refinements to planned outage modeling issues. Thank you for the opportunity to provide feedback on this important topic.