RASC: Planned Outage Modeling in LOLE (20210106)

MidAmerican Energy Company appreciates this opportunity to comment on the treatment of planned outages in Loss of Load Expectation studies.

MISO describes three options for treating planned outages:
• Outages scheduled with perfect foresight of load conditions (so called “optimal outages”)
• Outages scheduled according to average load conditions (so called “realistic outages”)
• Outages calibrated to historic schedules (so called “calibrated outages”)

MidAmerican would reject the first option. While either of the last two options may be useable, MidAmerican leans toward the third option (calibrated outages”) since it tends to result in a PRM between the two remaining options.

As an initial issue, we agree that the term “realistic outages” is unfortunate in that it implies a pre-determined preference for this option and implies that no other option is realistic.

MidAmerican believes that modeling outages with perfect foresight of statistically variable load conditions (so called “optimal outages”) is completely unsupportable.

Modeling outages based on average load conditions (so called “realistic outages”) aligns with how many Market Participants would schedule outages with long lead times. These prescheduled outages are difficult to move, because they typically involve extensive use of outside contractors. However, the use of average load conditions may be overly conservative in that it doesn’t recognize the ability to reschedule certain short term, short duration outages based upon load conditions.

The third option (so called “calibrated outages”), has the pragmatic advantage of setting a PRM between the extremes of the other options. It also appears to base the LOLE study on the conditions that MISO has actually observed. If MISO models outages on historic schedules, those “historic schedules” should reflect the post-RAN Phase I scheduling provisions rather than including older outages that RAN Phase I was specifically intended to alter. We realize that doesn’t give MISO much history to work with, and for that reason MISO may want to apply judgment in using this option.

It’s possible that a change in the modeling of planned outages will result in an abrupt change in PRM. If that occurs, MidAmerican believes that any change should be phased in over time rather than presenting Market Participants with sudden, volatile changes in requirements; Market Participants need time to make reasoned decisions and should not encounter new, unachievable requirements shortly before the Planning Year.

We note that some parties have called for MISO to exercise more authority in scheduling or approving generation outages in order to improve reliability, reduce the Planning Reserve Margin, etc. While MidAmerican doesn’t object to discussing the merits of more centralized coordination of planned outages in the long term, we don’t believe short term LOLE studies should be altered until MISO actually achieves a change in outage coordination. Short term LOLE studies should reflect the current conditions that MISO actually faces.

DTE appreciates this opportunity to provide feedback on modeling planned outages in the LOLE study.

DTE supports utilizing the calibrated outage modeling option for PY 22-23’s LOLE study assuming the work required to implement this change isn’t prohibitive. Based on the graphs shown in the January RASC meeting, this methodology appears to align best with how planned outages are occurring. There is a clear pattern over the past 3 years that will likely not change as long as the resource adequacy construct remains annual with a focus on the summer months. DTE seeks additional clarification on a couple of aspects of the calibrated approach:

• How many historic years would be included in the calibrated approach if selected?
• Can MISO provide data over a larger historical period so stakeholders can see if this trend in planned outages is consistent?

For the 2022-2023 Planning Year (PY), WEC Energy Group supports the continued use of the “hybrid” planned outage modeling used for the 2021-2022 PY LOLE study.  Use of the “realistic” or average planned outage modeling for the MISO-wide PRM analysis recognizes that planned maintenance is never optimal, especially on a system as geographically diverse as MISO.  However, on a zonal basis, we believe use of the optimal planned outage scheduling is appropriate because resource owners within each of the Local Resource Zones (LRZs) are more likely to coordinate with one another and consider Maintenance Margins within the zone.

We are interested in understanding the implications of the sub-optimal planned outage modeling, especially as applied on the LRZ level.  On a periodic basis, MISO and stakeholders should review the appropriateness of the LOLE planned outage modeling as we gain more experience with the RAN Phase 1 Outage Coordination requirements and associated tools such as the Maintenance Margin.

1. The LOLE outage modeling should reflect the same level of MISO outage coordination that is experienced in the MISO actual outage coordination. 
2. The Optimal modeling approach is stated as including “perfect foresight” of system load.  There is an expected dependency on any level of outage coordination and the expected availability of  wind and solar generation resources.  LOLE modeling should properly reflect the uncertainty in the dispatch of renewable generation that was experienced in the fall of 2020, when there were hours with significant unexpected decreases in the wind generation.  The underlying uncertainty of the wind and solar generation output needs to be understood, properly represented from a statistical distribution perspective in the LOLE analysis.

Alliant Energy prefers the “Optimal” outage modeling approach for both MISO footprint and zonal levels.  The “Realistic” and “Calibrated” methods are flawed in that they do not accommodate operational flexibility of rescheduling outages based on developing conditions.  

With MISO proposing “availability” accreditation, any modeling approach that does not offer the flexibility of the Optimal approach may result in double penalty between capacity requirements and capacity accreditation.  

MISO should allow for the potential that its ability to foresee tight capacity conditions will improve with time.

Alliant Energy prefers the “Optimal” outage modeling approach for both MISO footprint and zonal levels.  The “Realistic” and “Calibrated” methods are flawed in that they do not accommodate operational flexibility of rescheduling outages based on developing conditions. 

With MISO proposing “availability” accreditation, any modeling approach that does not offer the flexibility of the Optimal approach may result in double penalty between capacity requirements and capacity accreditation. 

MISO should allow for the potential that its ability to foresee tight capacity conditions will improve with time.

In addition to MGE and MPPA, SMMPA also generally supports WPPI Energy's feedback.

Sorry for the confusion.

David Sapper

dsapper@ces-ltd.com

MGE and MPPA generally support WPPI Energy's feedback.

Also, more informally, some LSE Coalition members have some preliminary concerns about using capacity accreditation in addition to generator outage assumptions in the LOLE Study for addressing resource adequacy needs and the potential for inefficiently and unfairly "double counting" outages.

Thanks,

David Sapper

dsapper@ces-ltd.com

The Entergy Operating Companies (“EOCs”) appreciate the opportunity to provide feedback on the LOLE enhancements discussed at the January 5, 2021 RASC meeting. During this meeting, MISO acknowledged concerns with the “realistic” outage scheduling approach, including an unreasonably sub-optimal scheduling of outages across seasons and complete inflexibility to reschedule outages in response to scarcity.  MISO also presented options to improve the modeling.

The EOCs appreciate that MISO has acknowledged these concerns and begun a stakeholder process to improve the modeling. In these comments, we describe another related issue – the modeling of import capacity – and propose modeling enhancements to address it.  These modeling enhancements should be adopted concurrently with the implementation of any planned outage scheduling approach, especially if such approach identifies non-summer LOLE risk.  Finally, we provide feedback on which of MISO’s proposed options would best address MISO and stakeholder concerns.

1. Treatment of Zonal Imports in LCR Calculation

The EOCs believe that the current treatment of Zonal Import Ability (“ZIA”) in the Local Clearing Requirement (“LCR”) calculation suffers from a mismatch between the timing of LOLE risk throughout the year and the capacity accredited to zonal imports. The LCR is currently calculated using a two-step process. First, the Local Reliability Requirement (“LRR”) is calculated by zone, ignoring any reliability benefit from zonal imports. Second, the LRR is reduced by the summer ZIA to calculate the LCR. This calculation suffers from a critical flaw in that it assumes that all LOLE risk occurs during the summer, and that only the summer ZIA value is relevant to reliability. Modeling done by MISO and observations of MaxGen events suggests that this assumption is not valid. In fact, MISO’s ongoing work to develop a sub-annual resource adequacy construct acknowledges this inconsistency. Since zonal imports may have higher ratings during winter and shoulder seasons, the EOCs believe the LCR calculation is biased high and does not accurately reflect zonal reliability needs.

2. EOC Proposal for Zonal Imports

The EOCs believe that the zonal LOLE/LCR calculation should be modified in the following way:

1. Each zonal LOLE model should include a single proxy resource that represents zonal imports. This resource should have perfect reliability and have seasonal MW ratings that match that zone’s summer, winter, and shoulder zonal import capability.
2. Incremental capacity should be added or subtracted as necessary to achieve LOLE=0.1.
3. The LRR will equal the existing zonal capacity plus the incremental capacity but will exclude the proxy unit representing zonal import capability.
4. The LCR will equal the LRR, since the reliability benefit of zonal imports will already have been accounted for in the LOLE model through the proxy resource.

The EOCs believe that this approach is conceptually sound, straightforward to implement, and is a more realistic representation of zonal reliability than the current methodology. This proposal specifically addresses the zonal LCR calculation. The proxy units described above would not be included as resources when MISO models systemwide LOLE risk. However, MISO should consider modeling external support at the systemwide level using seasonal ratings as well. This proposal is independent of the planned outage modeling options discussed at the January 5 RASC meeting, and the EOCs believe that MISO should adopt this approach regardless of how it ultimately chooses to model planned outages in its zonal and systemwide LOLE models, but especially if the zonal modeling identifies non-summer LOLE risk.

3. ZIA Should Assume Import Maximizing Re-Dispatch

The EOCs also believe that the methodology used to calculate ZIA is too conservative and should be modified. First, the re-dispatched base case that serves as the starting point for the transfer is sub-optimal due to constraints imposed on the redispatch.  Second, the transfer is proportional, which does not account for the impact of individual units on the import constraint. In practice, this does not accurately reflect system operation during scarcity conditions. We believe that calculating the ZIA by re-dispatching resources to maximize import capability would be straightforward to implement and would more closely match real-world operations than the current modeling approach.

4. Feedback on Planned Outage Modeling Proposals

The EOCs believe that, among the options proposed by MISO, Option 1 represents the best balance between ease of implementation and fidelity to the underlying system. Although imperfect, it does not suffer from the more serious deficits exhibited by Option 2.

As noted above, we believe that MISO should revise its treatment of ZIA to reflect both optimal re-dispatch methodology and seasonality, regardless of the option ultimately pursued. It is particularly important in the case of Option 2, as Option 2 could lead to serious inconsistencies in modeling assumptions where a significant amount of non-summer LOLE risk is present and is paired with summer import capability. This would result in an erroneously high LCR.

Further, as noted by MISO, Option 2 does not reflect real-world observations that planned outages can be, and are, modified in real-time to offset address conditions. Since Option 1 utilizes weather-year specific planned outage profiles, it does a better job at mimicking real-world system behavior in this respect.

The EOCs do not have enough information related to Option 3 at this time to fully evaluate the proposal. However, to the extent that Option 3 makes use of MISO’s "realistic" outage scheduling approach or any other approach that includes non-trivial non-summer LOLE risk, the EOCs note that it is especially important to incorporate the ZIA reforms described above in sections 2 and 3.

The OMS Resources Work Group (OMS RWG) acknowledges that improvements on the previously used optimal planned outage methodology are needed to accurately reflect how planned outages occur in reality. Of the choices listed by MISO in its January 6^th presentation, the OMS RWG believes that pursuing Option 2 or Option 3 would be the preferred path forward, as the approach in Option 1 was viewed as a stopgap. 

However, these options need to be reviewed and vetted thoroughly, with the impacts to both the region-wide Planning Reserve Margin and the individual LRZ Local Reliability Requirements known well before any final decision is to be made. In reviewing the options, the OMS RWG values stability in the reliability requirements moving forward, as significant changes year-to-year impose significant challenges on resource planning. 

The OMS RWG also supports the stakeholder comment about changing the “realistic” designation of the methodology that schedules planned outages according to average load conditions.

This feedback does not constitute a position of the OMS Board of Directors.

Xcel Energy believes that the planned outage modeling in the LOLE process needs to be representative of actual outage scheduling and account for the generator outage coordination process.  This process allows generators to schedule outages between 14 and 120 days in advance if the Maintenance Margin demonstrates that there is sufficient capacity available to serve the load.  Therefore, when looking back at historical data for calibration, any planned outage that was scheduled between 14 and 120 days in advance and was approved by MISO should not be incorporated into the LOLE process as the outage was taken because the net capacity margin was large enough to accommodate the outage.  One should assume that MISO's outage coordination process is robust enough to deny the outage if sufficient capacity margin does not exist and this should be reflected in the planned outage modeling methodology in the LOLE.  

Of the three options: (1) retain approach implemented in PY21/22 LOLE study, (2) calibrate outage schedules according to historic planned outage schedules, or (3) develop sub-optimal outage scheduling with flexibility to adjust level of optimality, WPPI recommends MISO and stakeholders pursue Option 3.  We also believe that significant focus on outage scheduling in each LRZ is warranted, to ensure that LRZ-specific LRR calculations are reasonable and that the new requirements are understood by stakeholders prior to implementation. As noted in the RASC presentation (20210106 RASC Item 03c LOLE Enhancements (Outage Modeling), Option 2 doesn’t capture the flexibility to reschedule outages and Option 1 uses Realistic Planned Outages which we believe is too conservative.