Curtailment: Prairie Ridge Can Hit P50 and Still Lose $1.18 Million of Revenue

September 17, 2026 · Dezzmond Team

Energy Markets Market Structure Curtailment

Prairie Ridge Solar produces exactly its P50 energy.

Nothing is wrong with the resource.

The grid still does not take every megawatt-hour.

Keep the same 200 MW Prairie Ridge used in the basis and hub-settled PPA articles.

Assume:

Available P50 generation                    438,000 MWh
Hub-settled PPA strike                       $45/MWh
Generation-weighted hub price                $40/MWh
Generation-weighted node price               $34/MWh
Node-to-hub basis                             -$6/MWh
Curtailment                                   6.0%

The curtailed volume is:

438,000 × 6%
= 26,280 MWh

Delivered generation becomes:

438,000 - 26,280
= 411,720 MWh

Under the simple matched-volume hub-settled PPA from the prior article, each delivered MWh realizes:

Strike + basis
= $45 - $6
= $39/MWh

So Prairie Ridge earns:

411,720 MWh × $39
= $16.057m

The project hit P50.

Revenue still lost more than a million dollars of value because some of the available energy never reached the settlement stack.

Curtailment is a volume event

Basis and curtailment often have the same grid cause.

They are not the same financial variable.

Basis changes the price received on a delivered MWh.

Curtailment removes the MWh.

The clean model order is:

Available generation
        ↓
less curtailed MWh
        ↓
Delivered generation
        ↓
apply nodal / PPA settlement prices
        ↓
Revenue

Not:

P50 generation
× price haircut
× curtailment haircut

The order matters because the second approach can charge the same bad hour twice.

The double-count is $157,680 in the Prairie Ridge example

Start from the strike-value case:

438,000 MWh × $45
= $19.710m

The correct bridge is:

Curtailment first

26,280 curtailed MWh × $45
= $1.1826m volume loss

Basis only on delivered MWh

411,720 delivered MWh × $6
= $2.47032m basis loss

Net revenue

$19.710m
- 1.1826m
- 2.47032m
= $16.05708m

Now look at the common wrong model.

It applies basis to all P50 MWh:

438,000 × $6
= $2.628m

and then separately deducts curtailed energy at the strike:

26,280 × $45
= $1.1826m

That produces:

$19.710m
- 2.628m
- 1.1826m
= $15.8994m

The model is too low by:

$16.05708m - $15.8994m
= $157,680

Exactly:

26,280 curtailed MWh × $6/MWh basis
= $157,680

The model applied node-to-hub basis to energy that was never delivered.

That is the double count.

CAISO's own reporting separates cause and dispatch mechanism

California ISO's market reporting is useful because it does not treat all curtailed renewable MWh as one category.

CAISO reports Variable Energy Resource curtailment by:

economic curtailment;
self-schedule curtailment;
exceptional dispatch curtailment; and
whether the underlying reason is system-wide conditions or local congestion.

CAISO defines economic curtailment as the difference between the resource's dispatch upper limit and its real-time dispatch schedule when the resource has an economic bid.

Self-schedule curtailment is the difference between self-scheduled output and the lower real-time dispatch schedule.

That distinction is useful for finance.

The meter can show the same result:

available MWh
>
delivered MWh

while the commercial cause is different.

CAISO's April 2026 data show the scale can be very large

In April 2026, CAISO reported approximately:

Solar energy curtailed                 1,305,200 MWh
Wind energy curtailed                    110,300 MWh
                                       -------------
Wind + solar curtailed                 1,415,500 MWh

Solar therefore represented roughly:

1,305,200 / 1,415,500
≈ 92%

of the reported wind-and-solar curtailed energy that month.

That does not mean every California solar project should put a 92% curtailment haircut in its model.

It means system conditions can create very large aggregate curtailed volumes and that solar dominates the California renewable-curtailment profile in the spring.

Project finance still needs the node, local transmission, operating profile and contractual allocation.

System data is context.

It is not a project assumption.

ERCOT can show curtailment even when renewable output is setting records

ERCOT's June 21, 2025 renewable-record publication is a good example.

At 13:00, ERCOT reported:

Wind + solar generation                   46,966 MW
Renewable penetration                         64%
Wind / solar curtailment                    8,422 MW

The system was simultaneously producing record renewable output and curtailing more than eight gigawatts of additional wind and solar capability.

That is an important mental model.

Curtailment does not require:

renewables are not generating.

It can happen precisely when the renewable fleet is generating extremely strongly.

For a project in an export-constrained area, a strong regional resource hour can be the hour of highest curtailment risk.

Congestion and economic curtailment can be two stages of one mechanism

Prairie Ridge's node-to-hub basis example already showed the price effect of congestion.

Now deepen the same constraint.

Stage 1 — node still positive

Hub      $40
Node     $34
Basis    -$6

Prairie Ridge still generates.

The grid problem appears as price.

Stage 2 — node approaches zero or becomes negative

The same export constraint deepens.

The project now faces a dispatch / operating decision.

Depending on:

market dispatch;
PPA terms;
tax credits;
REC value;
operating cost;
negative-price provisions; and
curtailment instructions,

some MWh may no longer be economic or permitted to generate.

The grid problem now appears as volume.

That is why basis and curtailment are correlated.

It is also why multiplying an independently forecast basis haircut by an independently forecast curtailment haircut can be wrong unless the underlying hourly simulation reconciles them.

Economic curtailment is not the same as a reliability instruction

Suppose Prairie Ridge sees:

Node price                         -$35/MWh

The project can decide that generating is no longer economic under its actual contract and tax position.

That is an economic response.

Now compare a separate event:

ISO / utility instruction
→ reduce output by 50 MW

The project can be curtailed even if its private economics would have favored continuing to generate.

Those events can produce the same missing MWh.

They need different model flags.

I would keep:

Economic self-curtailment
Grid / transmission curtailment
Buyer-directed curtailment
Plant outage / availability loss

separate.

An outage is not curtailment.

A model that combines them loses both contractual and operational information.

Buyer-directed curtailment is a contract question

The grid does not decide whether a buyer owes Prairie Ridge deemed energy.

The PPA does.

Use an illustrative contract rule.

Assume Prairie Ridge's 26,280 curtailed MWh break down as:

Grid / congestion curtailment              15,280 MWh
Economic self-curtailment                    5,000 MWh
Buyer-directed curtailment                   6,000 MWh
                                           -----------
Total                                       26,280 MWh

Now assume, purely for teaching:

grid / congestion curtailment is uncompensated;
economic self-curtailment is uncompensated; and
buyer-directed curtailment earns $45/MWh of deemed-energy compensation.

Then the contract restores:

6,000 MWh × $45
= $270,000

Prairie Ridge's revenue becomes:

Delivered energy + PPA cash               $16.057m
Illustrative deemed-energy payment          0.270m
                                           -------
Total                                      $16.327m

The $270,000 is not a generic PPA rule.

It exists only because we inserted a specific compensation clause.

That is exactly how the project model should work.

Do not model "curtailment compensation = yes/no."

Model the defined cause, allowance, formula and excluded events.

Free curtailment allowances create a threshold, not a smooth haircut

Now add another illustrative PPA term.

Assume the buyer can direct:

4,000 MWh/year

without deemed-energy compensation.

Prairie Ridge experiences 6,000 MWh of buyer-directed curtailment.

Only:

6,000 - 4,000
= 2,000 MWh

are compensated.

At $45/MWh:

2,000 × $45
= $90,000

The economic consequence is discontinuous.

The first 4,000 MWh produce no contractual payment.

The next MWh does.

So the model should not use:

Buyer curtailment × average compensation %

unless that percentage was derived from the actual allowance mechanics.

The PPA creates a threshold.

Model the threshold.

PTC projects can have a different economic shut-down point

For a project earning a production-based tax credit, one delivered MWh can create value outside the power price.

The simplified generation decision can include:

Nodal power value
+ PTC value
+ REC / environmental value
+ PPA settlement effect
- variable operating cost
- other marginal costs

So a negative nodal price does not automatically mean curtailment is optimal.

And a zero-price floor in a PPA can change the incentive again.

This is why the hub-settled PPA article refused to hard-code:

negative price
→ stop

The curtailment model should use the same marginal-economics logic.

P90 generation does not include grid curtailment unless the energy case explicitly says it does

This is one of the most dangerous shortcuts in lender models.

A P90 energy-yield case is typically a resource / performance probability case.

It does not automatically mean:

90% probability after all future grid constraints, curtailment and commercial dispatch.

If the independent engineer's P90 already includes a stated curtailment assumption, then do not deduct it again.

If it is a gross or pre-curtailment energy case, the grid haircut still needs its own line.

The diligence question is:

Does P50 / P90 mean:
gross available energy?
net plant energy before grid curtailment?
or
net delivered energy after modeled curtailment?

Do not infer the answer from the label.

Read the energy report.

Strong resource and high curtailment can occur together

Suppose Prairie Ridge's region has an unusually sunny year.

The project's available energy is above P50.

So are the neighboring projects' outputs.

If the local export constraint is binding, the same strong resource condition can increase:

nodal oversupply;
negative basis;
negative-price hours; and
curtailed MWh.

That means resource uncertainty and curtailment are not necessarily independent.

A simplistic downside case can accidentally assume:

P90 generation
+
high curtailment

without checking whether that combination is physically coherent.

Or it can assume:

strong P50/P10 resource
+
normal curtailment

when strong regional output is exactly what worsens congestion.

The best model is hourly and correlated.

If the project model is annual, at least make the correlation assumption explicit.

Storage can change the host's curtailment without making the constraint disappear

Add a co-located battery behind Prairie Ridge's POI.

Assume the POI export limit remains 200 MW.

At noon:

Available solar                         200 MW
Grid / contract export opportunity      150 MW
Potential curtailment                    50 MW

If a battery can charge 50 MW behind the POI:

Solar generation                        200 MW
Export                                  150 MW
Battery charging                         50 MW
Curtailment                               0 MW

The battery did not increase the export capability of the line.

It absorbed energy that could not be exported in that interval.

Later it may discharge when:

solar output falls;
the node improves;
the export constraint clears; or
the combined project has spare POI headroom.

This can improve host revenue.

It still depends on:

interconnection rights;
grid-charging rules;
SOC;
battery efficiency;
degradation;
future price; and
the same POI limit.

The battery article and Fox Run interconnection series matter here: a financial dispatch optimization is only useful if the studied operating envelope permits it.

The host benefit must not be double-counted with the battery's own arbitrage

Suppose the battery captures 50 MWh of solar that otherwise would have been curtailed.

Later it discharges 44 MWh after losses.

There are two ways to make the model wrong.

Wrong approach 1

Give the solar project full "avoided curtailment revenue" on the 50 MWh and give the battery full arbitrage revenue as if it bought the 50 MWh at zero cost.

Wrong approach 2

Treat the 50 MWh as lost solar generation and still let the battery discharge energy from it.

The transfer of energy from host to battery needs an internal economic treatment.

A clean hybrid model can show:

Host solar opportunity cost / transfer value
Battery charging energy
Battery losses
Battery discharge revenue
Net hybrid value

The consolidated project should only count the MWh once.

Transmission relief can improve both volume and price

Assume a network upgrade reduces Prairie Ridge's curtailment from:

6% → 2%

and improves generation-weighted basis from:

-$6/MWh → -$3/MWh

Available generation stays:

438,000 MWh

Before upgrade

Delivered MWh
= 438,000 × 94%
= 411,720

Realized price
= $45 - $6
= $39

Revenue
= $16.057m

After upgrade

Delivered MWh
= 438,000 × 98%
= 429,240

Realized price
= $45 - $3
= $42

Revenue
= 429,240 × $42
= $18.028m

Revenue improvement:

$18.028m - $16.057m
≈ $1.971m/year

Part came from more MWh.

Part came from a better price on every delivered MWh.

That is why a transmission upgrade can be valuable even when the curtailment percentage alone looks modest.

Decompose the $1.971 million improvement

Use the old delivered volume and the new price first.

Price effect

411,720 MWh × ($42 - $39)
= $1.235m

Volume effect

New delivered volume:

429,240 - 411,720
= 17,520 MWh

Value at new $42 realized price:

17,520 × $42
= $0.736m

Total:

$1.235m + $0.736m
= $1.971m

Roughly:

63% price / basis improvement
37% volume / curtailment improvement

in this teaching case.

The grid upgrade's value would be badly understated if the model measured only avoided curtailed MWh.

The financing sensitivity should use delivered energy and capture price separately

Prairie Ridge's lender should not ask for one generic "grid risk" haircut.

Keep at least two variables.

Curtailment	Basis	Delivered MWh	Realized price	Revenue
2%	-$3	429,240	$42	$18.028m
2%	-$6	429,240	$39	$16.740m
6%	-$3	411,720	$42	$17.292m
6%	-$6	411,720	$39	$16.057m
10%	-$8	394,200	$37	$14.585m

Now the credit team can see whether the downside is driven by:

fewer MWh;
lower price;
or both.

A single "capture rate" can be useful for a merchant model.

It is less useful for diagnosing what covenant, contract or grid remedy changes the answer.

The PPA and merchant model need the same curtailment taxonomy

For Prairie Ridge I would use one event table.

Event	MWh delivered?	Physical market revenue?	PPA settlement?	Deemed compensation?
Economic self-curtailment	No	No	Contract-specific	Usually model from contract
Grid / congestion curtailment	No	No	Contract-specific	Contract-specific
Buyer-directed curtailment	No	No	Contract-specific	Contract-specific
Plant outage	No	No	Availability / PPA treatment	Not "curtailment" by default
Normal generation	Yes	Yes	Yes per PPA formula	No

The exact cells depend on the actual document.

The taxonomy should not.

Without it, the same MWh can be:

removed from generation;
charged basis;
denied PPA settlement; and
denied deemed energy

four different times.

What the diligence team should request

For Prairie Ridge:

Is the energy report gross or net of historical / forecast curtailment?
How does the market / consultant define curtailment?
Is the forecast node-specific or based on a system average?
Does the model separate local congestion from broader oversupply?
Are curtailed MWh removed before applying nodal basis?
If a capture price or capture rate is already used, what grid effects are embedded in it?
Does the PPA compensate any buyer-directed or other defined curtailment?
Is there a free curtailment allowance?
How is deemed energy calculated?
Do negative prices change the seller's generation obligation?
Does PTC or REC value change the economic shut-down threshold?
Does co-located storage truly reduce curtailed host MWh under the actual interconnection envelope?
Is the battery's host benefit reconciled with its own charging cost and arbitrage revenue?
Which funded transmission upgrade is assumed to improve curtailment or basis?
What happens if that upgrade is delayed?

Prairie Ridge can hit P50 and still miss its revenue case.

The key is not to call that "underperformance."

The resource delivered what the resource model promised.

The grid accepted less of it.

And the revenue model has to know exactly which MWh disappeared before it prices the ones that remained.

Primary sources: CAISO — April 2026 Renewable Resource Market Performance · CAISO — Monthly Renewables Performance Report, April 2026 · CAISO — Renewable Curtailment Report Library · ERCOT — June 21, 2025 Renewable Generation and Curtailment Record · FERC — Understanding Energy Markets