PJM ELCC: A 100 MW Four-Hour Battery Can Be 59 MW of Capacity — Then 19 MW
Keystone Battery is a 100 MW / 400 MWh storage project in PJM.
For the 2028/2029 Base Residual Auction, PJM's final ELCC Class Rating for four-hour storage is 59%.
Assume for the moment:
Effective nameplate capacity 100 MW
ELCC Class Rating 59%
Resource Performance Adjustment 1.00
Capacity Interconnection Rights 100 MW
The first-pass calculation is:
100 MW × 59% × 1.00
= 59 MW
Because Keystone's 100 MW of CIRs are above that amount, the CIR cap does not bind.
Keystone therefore has 59 MW of Accredited UCAP in this simplified example.
At the actual 2028/2029 PJM capacity clearing price of $325/MW-day:
59 MW
× $325/MW-day
× 365
=
$6.999m annual capacity revenue
The battery is still a 100 MW battery.
Its capacity-market quantity is 59 MW.
That distinction is ELCC.
The class rating is not the project's final capacity value
A common model uses:
Nameplate MW × ELCC Class Rating
and stops.
Current PJM rules contain another layer.
For Variable Resources and Limited Duration Resources, Accredited UCAP is generally the lesser of:
Capacity Interconnection Rights
and:
Effective Nameplate Capacity
× ELCC Class Rating
× ELCC Resource Performance Adjustment
For Keystone, we assumed the Performance Adjustment is 1.00.
Now change it to 0.90.
100 MW × 59% × 0.90
= 53.1 MW
The same 59% class rating now produces only 53.1 MW of Accredited UCAP.
Now assume instead that Keystone has only 50 MW of applicable Capacity Interconnection Rights.
ELCC formula result 53.1 MW
CIR limit 50.0 MW
-------
Accredited UCAP 50.0 MW
Three quantities therefore belong in separate model cells:
Class Rating
Performance Adjustment
CIR / deliverability limit
The class rating is important.
It is not the final answer.
What the class rating is measuring
ELCC is not a capacity-factor test.
PJM's governing rules define the class rating through reliability contribution: the expected-unserved-energy improvement from adding an incremental quantity of the resource class, compared with the improvement from adding an incremental perfectly available unlimited resource.
That is why a resource can produce a large amount of annual energy and receive a low capacity rating.
The question is not:
How many MWh did the asset generate?
It is:
How much did this class reduce system reliability risk during the hours in which load loss matters?
That also explains why the number moves when the rest of the system changes.
ELCC is a portfolio reliability value, not a permanent physical characteristic stamped onto the plant.
Final and preliminary ratings are different inputs
This distinction matters enough to give it its own schedule.
PJM publishes final class ratings for specific capacity auctions.
It also publishes preliminary class ratings for future delivery years.
The latest September 2026 PJM forward table is explicit:
- 2028/2029 and 2029/2030 are the ratings used or to be used for those respective BRAs;
- later years are non-binding and indicative; and
- the forward resource portfolios for those later years depend on forecast additions and retirements.
For four-hour storage:
| Delivery year | PJM class rating | Status |
|---|---|---|
| 2028/29 | 59% | Final BRA rating |
| 2029/30 | 57% | Final BRA rating |
| 2030/31 | 42% | Preliminary / non-binding |
| 2031/32 | 34% | Preliminary / non-binding |
| 2032/33 | 30% | Preliminary / non-binding |
| 2033/34 | 26% | Preliminary / non-binding |
| 2034/35 | 24% | Preliminary / non-binding |
| 2035/36 | 23% | Preliminary / non-binding |
| 2042/43 | 19% | Preliminary / non-binding |
A financing model should not label the 2042/43 figure “PJM ELCC assumption” as though PJM has committed to it.
It should label it:
Current PJM preliminary forward class rating
Non-binding
As of Sept. 2026
That is accurate and still useful.
Hold the capacity price constant and isolate the accreditation effect
The 2029/2030 capacity auction has not yet produced a clearing price as of October 5, 2026.
So do not mix a future class rating with an invented future auction result.
Instead, hold $325/MW-day constant only as a sensitivity to isolate what the accreditation curve does.
Assume:
Performance Adjustment 1.00
CIRs not binding
Illustrative price $325/MW-day in every row
Then Keystone's capacity revenue would be:
| Rating | Accredited MW | Illustrative annual revenue at $325 |
|---|---|---|
| 59% | 59 MW | $6.999m |
| 57% | 57 MW | $6.762m |
| 42% | 42 MW | $4.982m |
| 34% | 34 MW | $4.033m |
| 30% | 30 MW | $3.559m |
| 23% | 23 MW | $2.728m |
| 19% | 19 MW | $2.254m |
From 59% to 19%:
Accredited MW decline 40 MW
Illustrative revenue decline
= $6.999m - $2.254m
= $4.745m per year
That is a 68% reduction in the modeled capacity line without changing:
- Keystone's 100 MW inverter rating;
- its four-hour duration;
- the assumed capacity price; or
- the project's physical construction.
The system value changed.
The project did not.
Do not turn the preliminary curve into a deterministic decline
The old version of this article made a second mistake: it treated declining accreditation almost as a one-way law.
PJM's current published data show why that is too simple.
The final onshore-wind class rating rises from:
2028/29 34%
2029/30 37%
And PJM's latest preliminary forward table then shows:
2030/31 40%
2031/32 41%
before declining later in the horizon.
Tracking solar is also not a smooth downward staircase:
2028/29 10%
2029/30 7%
2030/31 8% preliminary
2031/32 8% preliminary
ELCC can move in either direction when:
- the assumed resource mix changes;
- load shape changes;
- retirements change the risk hours;
- resource performance changes;
- outage assumptions change; or
- the reliability-risk distribution moves between seasons.
A good model can carry a declining central case.
It should not claim that decline is mechanically guaranteed.
Why four-hour storage falls so sharply in the current forward study
PJM's 2025 ELCC / Reserve Requirement Study gives a more useful explanation than “too much storage.”
The study points to increasing winter reliability risk.
For storage, the issue includes:
- longer expected winter reliability events;
- energy-constrained winter days;
- the ability to recharge between periods of system stress; and
- the model treatment of storage charging during constrained conditions.
A four-hour battery is highly valuable if the critical shortage lasts four hours and the system can recharge it before the next event.
It is less valuable if the event is longer or the charging opportunity disappears.
That is why duration matters.
For the final 2029/2030 BRA ratings:
| Storage duration | Final ELCC Class Rating |
|---|---|
| 4-hour | 57% |
| 6-hour | 66% |
| 8-hour | 69% |
| 10-hour | 76% |
Keystone at four hours receives 57%.
A 100 MW ten-hour battery in the same class-rating table starts from 76% before its own Performance Adjustment and CIR limit are applied.
That 19-point difference is the capacity market assigning value to duration.
The value of adding duration can be translated into dollars
Keep the price at $325/MW-day only for an apples-to-apples sensitivity.
Assume both resources:
100 MW nameplate
Performance Adjustment = 1.00
CIRs not binding
Four-hour battery
100 MW × 57%
= 57 MW
57 × $325 × 365
= $6.762m
Ten-hour battery
100 MW × 76%
= 76 MW
76 × $325 × 365
= $9.016m
Difference:
19 MW of Accredited UCAP
$2.254m annual illustrative capacity revenue
That does not prove an extra six hours of storage is economic.
The extra duration has capex, efficiency, degradation and dispatch consequences.
But it gives the investment model a clean reliability-value line to compare with that incremental cost.
The relevant question becomes:
What is the NPV of the extra accredited capacity, under a defensible capacity-price forecast, relative to the capex and operating cost of the longer-duration configuration?
That is a capital-allocation question.
ELCC is one input to it.
Solar shows why nameplate is a poor capacity-market shortcut
Return to the 200 MW tracking-solar project from ERCOT vs PJM.
The final 2028/2029 tracking-solar class rating is 10%.
Assume for illustration:
Effective nameplate 200 MW
Class Rating 10%
Performance Adjustment 1.00
CIRs not binding
Accredited UCAP:
200 × 10%
= 20 MW
At the actual 2028/2029 $325/MW-day price:
20 × $325 × 365
= $2.373m
For the 2029/2030 BRA, the final tracking-solar class rating is 7%.
Holding the price at $325 only to isolate the rating effect:
200 × 7%
= 14 MW
14 × $325 × 365
= $1.661m
Difference:
6 MW less accredited capacity
$0.712m less annual revenue at the same price
Again, the actual 2029/2030 auction price is not yet known.
The point is to separate:
Accreditation effect
from
capacity-price effect
Otherwise a model can show a single capacity-revenue decline and never reveal which risk caused it.
The Performance Adjustment makes the project matter again
Class ratings are system-level values.
PJM does not simply apply the same final percentage to every project and stop.
The ELCC Resource Performance Adjustment can move a resource above or below its class average based on the PJM methodology and the resource's expected performance during modeled loss-of-load-risk hours.
For Keystone:
Class rating 59%
Performance Adjustment 0.90
Effective class contribution 53.1% of nameplate
Another four-hour battery with a 1.05 adjustment would produce:
100 MW × 59% × 1.05
= 61.95 MW
before any CIR cap.
So the sentence:
Every four-hour battery gets 59 MW per 100 MW
is wrong.
The 59% rating is the class input.
The project's accredited amount also depends on the project-specific adjustment and deliverability limit.
Capacity Interconnection Rights can become the binding variable
Now assume Keystone has:
Nameplate 100 MW
Class Rating 59%
Performance Adjustment 1.00
Formula result 59 MW
CIRs 45 MW
Accredited UCAP is:
MIN(59, 45)
= 45 MW
The ELCC analysis says 59.
The project's interconnection rights say 45.
The market quantity is 45.
That is why the financial model should not infer capacity rights from nameplate.
The applicable CIR quantity belongs in the diligence file.
It is a separate project right created through the interconnection framework.
Hybrid resources can require a different accreditation path
Another old shortcut is to assume:
Solar class rating
+
battery class rating
=
hybrid capacity value
PJM's current rules do not support that generic arithmetic for every hybrid.
Certain Combination Resources, Complex Hybrid resources and other classes can receive resource-specific ELCC analysis rather than a single published class rating.
That matters for co-located solar-plus-storage projects.
The project configuration, charging restrictions, interconnection rights and operating constraints can affect the resource-specific answer.
Do not value a hybrid by mechanically adding:
200 MW solar × 7%
+
100 MW battery × 57%
unless the applicable PJM classification and accreditation rules actually allow that calculation for the specific project.
The market's definition of the resource comes before the spreadsheet formula.
Final ratings can change after the auction commitment
There is another financing wrinkle.
The capacity obligation and the eventual accredited quantity do not necessarily live on a frozen number forever.
PJM's tariff includes Capacity Resource Deficiency Charge mechanics where a resource's Accredited UCAP Factor is reduced between the Base Residual Auction and the start of the Delivery Year because of changes in ELCC Class Ratings or Performance Adjustments.
That means accreditation is not merely a forecasting input.
It can interact with an already-created capacity commitment.
A project that offers too aggressively against an expected rating can have a replacement-capacity or deficiency-cost problem if the later accredited amount falls.
For financing, I would therefore keep:
Auction commitment MW
Latest Accredited UCAP
Replacement / deficiency exposure
on the same schedule.
The lender cares about the net cash after the obligation, not only the gross capacity award.
The fifteen-year preliminary curve is useful precisely because it is not a promise
PJM extended its September 2026 preliminary class-rating horizon through 2042/2043 to align with the 15-year term of the Reliability Backstop Procurement framework.
That makes the data unusually useful for project finance.
It gives Keystone a published view across something close to a debt tenor.
But PJM also says explicitly that the later-year values are non-binding and indicative.
That combination is exactly how they should be used:
not as contracted revenue,
and not ignored.
For Keystone I would build at least three cases.
Case 1 — Current final rating held flat
57% after 2029/30
Useful as a simple reference case.
Not a neutral forecast.
Case 2 — Current PJM preliminary curve
57% → 42% → 34% → ... → 19%
Useful as the current official indicative planning case.
Still not binding.
Case 3 — Financing downside
A separate lender / sponsor downside around:
- faster rating decline;
- lower Performance Adjustment;
- lower CIR availability;
- lower capacity price; and
- replacement / deficiency costs.
The model should never hide all five risks inside one “capacity haircut.”
What the lender should actually size
Keystone's capacity revenue is a product of multiple variables:
Accredited UCAP
=
MIN(
CIRs,
Effective Nameplate
× Class Rating
× Performance Adjustment
)
Capacity Revenue
=
Cleared / contracted capacity MW
× Capacity Price
× Delivery Period
- penalties / deficiency / replacement costs
Those are separate lines because they change for different reasons.
A lender can choose to credit:
- final ratings more heavily;
- preliminary ratings with haircuts;
- only a subset of forward capacity revenue;
- or no merchant capacity revenue outside a supported horizon.
The important thing is that the haircut be applied to the correct variable.
A 50% haircut to capacity price is not economically identical to a 50% haircut to accredited MW if commitment obligations or CIR limits also matter.
Keystone's capacity page
A clean model should display the capacity build before it displays revenue.
Resource facts
Nameplate capacity 100 MW
Storage duration 4 hr
CIRs 100 MW
Performance Adjustment 1.00
Rating status
2028/29 final class rating 59%
2029/30 final class rating 57%
2030/31 onward
current preliminary / non-binding curve
Accredited UCAP
MIN(
CIRs,
Nameplate × Rating × Performance Adjustment
)
Revenue
Accredited / committed MW
× $/MW-day
× days
=
gross capacity revenue
Risk bridge
Class-rating movement
Performance-adjustment movement
CIR / deliverability cap
Capacity-price movement
Deficiency / replacement exposure
=
net capacity variance
If the model reports only:
Capacity revenue = $6.999m
it has hidden the information the financing team needs most.
The diligence questions are now very specific
For Keystone:
- What is the resource's PJM ELCC class?
- Is the published value final for the relevant BRA or preliminary for a future year?
- What publication date and delivery year does the model use?
- What ELCC Resource Performance Adjustment applies to the project?
- What Effective Nameplate Capacity is used in the formula?
- What Capacity Interconnection Rights cap the result?
- Is the resource class-rated or subject to resource-specific ELCC analysis?
- For storage, what duration and charging constraints are represented?
- For a hybrid, what PJM classification applies to the actual configuration?
- What capacity MW have already been committed?
- Can a later accreditation reduction create deficiency or replacement costs?
- What capacity price is cleared, what price is only a rule parameter, and what price is purely a forecast?
- Does the debt model use the current preliminary ELCC curve as information or mistakenly treat it as contracted capacity?
Keystone is physically the same 100 MW / 400 MWh battery in every row of the model.
Its final 2028/2029 class input is 59%.
The next final rating is 57%.
The current non-binding forward view reaches 19% by 2042/2043.
That does not mean PJM has promised a decline to 19%.
It means the current reliability model is telling the financing model something valuable: do not treat today's accredited MW as a permanent physical attribute of the asset.
Primary sources: PJM — ELCC Process and Published Results · PJM — Final 2028/2029 ELCC Class Ratings · PJM — Final 2029/2030 ELCC Class Ratings · PJM — Preliminary ELCC Class Ratings 2028/29–2042/43 · PJM — 2025 ELCC / Reserve Requirement Study · PJM RAA — Accredited UCAP and Performance Adjustment · PJM — 2028/2029 Capacity Auction Result