Who carries the risk? Power purchase agreements and risk allocation

Sign four contracts for a solar-plus-storage plant, run 200 price years, and see where the risk lands

Analyst lesson A2 of Energy Systems, Step by Step: sign fixed-price, hub-settled and shaped contracts and a battery toll, simulate 200 price years, and see who carries which risk.
  • Analyst · L2
  • 25 min guided · about 30 min practice
  • After A1

One plant, four contracts, 200 simulated price years (seed 2025; real 2025 dollars)

  1. A 100 MW solar farm makes 210,240 MWh in a typical year; a buyer needs the same amount, flat. Five drivers change each year (first table below).
  2. With no contract, the solar farm's sales run from $3.9M (P10) to $8.2M (P90) a year, and the buyer's bill from $5.2M to $10.2M.
  3. A fixed price settled at the node leaves the seller only output risk ($5.8M to $6.3M). The buyer carries price, shape and basis, and shares output risk: the settled MWh follow output ($6.8M to $8.6M).
  4. Settled at the hub, the basis moves to the seller ($5.3M to $6.7M); the buyer's range narrows ($7.0M to $8.3M).
  5. A shaped contract, a flat block at a fixed price, fixes the buyer's bill at $7.59M; the seller settles the hours its panels miss too ($5.0M to $6.9M).
  6. Tolled at $5.17 per kW a month, the battery pays its owner a fixed $3.10M a year, and its margin ($1.9M to $4.5M) moves to the buyer.
  7. A price $5.00 above fair moves $1.05M a year from the buyer to the seller; who carries which risk stays the same.
Driver What it is
Hub price The yearly average price at the trading hub, a regional reference price
Shape The hub price per MWh of solar output, as a share of the hub's all-hours average (the capture ratio)
Output This year's solar output, as a share of a typical year's
Basis The plant's node price minus the hub price, per MWh of solar output
Battery margin What the battery earns per MWh after paying for its charging and losses
Contract Fair price Seller's sales, P10 to P90 Buyer's bill, P10 to P90
No contract – $3.9M to $8.2M $5.2M to $10.2M
Fixed price, settled at the node $28.69/MWh $5.8M to $6.3M $6.8M to $8.6M
Fixed price, settled at the hub $30.69/MWh $5.3M to $6.7M $7.0M to $8.3M
Fixed price, shaped: a flat block $36.11/MWh $5.0M to $6.9M $7.59M, every year

With JavaScript on, you draw price years, sign each contract and see both parties' ranges move, with a seed you can change.

  • A teaching simulation: the plant, the buyer and the five drivers' spreads are illustrative, and the drivers are drawn independently from a seeded generator (seed 2025 here).
  • A price year is one simulated year, not a calendar year and not a forecast.
  • Every contract here is settled in cash, as a financial PPA: the buyer buys its power at the hub, and the contract pays the price difference.
  • P10 and P90 are statistics of the 200 simulated years: one year in ten fell below the P10 and one in ten above the P90 (percentiles as in toolbox K5).
  • Solar capacity factor 24.0%: EIA, Electric Power Annual, Table 4.08.B (U.S. utility-scale solar PV generators, 2025; October 2026), as in lesson 7.
  • One price year: all dollars are 2025 dollars.
  • Left out: taxes, a party failing to pay, curtailment, capacity payments, renewable certificates, the battery's wear, and the hours inside each year.

Need a refresher? A1 (a project's cash flows) and toolbox K5 (percentiles and sampling).

What you just did, as diagrams

Power purchase agreement
A long contract to buy a plant's power, or its value, at a set price.
As generated
The contract covers whatever the plant makes: the seller is paid on its actual output, and the settled MWh move with it, so the buyer feels output too.
Settled at the node or the hub
The price difference is paid against the plant's own node or a trading hub; the basis between them sits with whoever is not paid at the node.
Shaped
A fixed amount every hour, settled in cash: the seller owes the difference in the hours its panels miss, so it carries the shape.
Toll
The buyer pays a fixed fee to run the plant, here the battery, and keeps what it earns.
P10 to P90
The middle 80% of the simulated years: a stated spread, not the worst or best case.
Go deeper · AnalystWhy contracts matter to lenders, and a naming trap
  • Lenders size loans on the bad years. A loan is repaid from the low years’ revenue, so a narrower range can carry a larger loan. This stays at awareness level here (side shelf X5).
  • Fair is not the market price. Each contract here is signed at its fair price, so on average it moves no money. Real prices also pay one side for the risk it takes on.
  • A naming trap. Resource studies often name levels by exceedance: their “P90” output is the level exceeded in 90% of years, this page’s P10. Check which convention a study uses.
Go deeper · ModelerSeeds, sampling noise and linked drivers
  • Seeded, so reproducible. The same seed gives the same 200 years on any computer. A new seed gives other years, and the P10 and P90 move a little: that is sampling noise (K5’s standard error).
  • Independence is an assumption. Here the five drivers are drawn independently. In real markets a high-price year may also be a wide-spread year for the battery; linked inputs and global sensitivity are Modeler lesson M7.
  • Hours inside a year. Each driver is one number per year. An hourly model would settle each hour, and lets the battery cover a shaped contract’s evening hours.
  • Cash or delivery. Every contract here is settled in cash. As physical delivery, the seller buying the hours its panels miss, the flat block gives the same money only if bought at the hub.

Check yourself

Myth or fact? “A fixed-price contract at a fair price makes the seller richer.”

  • Myth
  • Fact
Answer

Myth. It moves risk, not money: over the 200 years the solar farm averaged $5.92M a year with no contract and $6.01M settled at the hub, a gap of sampling noise.

Myth or fact? “Under a hub-settled fixed price, the seller carries no risk.”

  • Myth
  • Fact
Answer

Myth. Output and basis stay with the seller: its sales still run from $5.3M to $6.7M (P10 to P90).

Myth or fact? “Under a flat block at a fixed price, the buyer's bill no longer depends on the hub price.”

  • Myth
  • Fact
Answer

Fact, in this model: the buyer pays $7.59M every year, and the seller carries the block's price and shape risk.

Myth or fact? “The P10 is the worst case.”

  • Myth
  • Fact
Answer

Myth. With no contract, 20 of the 200 years fell below the P10 of $3.9M; the lowest was $2.4M.

Seed 2025, price year 1: the hub averaged $68.14/MWh, the highest of the 200. Over the solar farm's 210,089 MWh it averaged $54.47/MWh. Under the hub-settled fixed price of $30.69/MWh, what happens?

  • The buyer pays the seller $5.00M
  • Nobody pays: the price is fixed
  • The seller pays the buyer $5.00M
  • The seller keeps the high price
Answer

The seller pays the buyer $5.00M: 210,089 × (30.69 − 54.47) ≈ −5.00 million dollars. The hub price per solar MWh is above the fixed price, so the contract pays the difference to the buyer.

Which contract leaves the buyer's bill the same in every price year of this model?

  • No contract
  • Fixed price, settled at the node
  • Fixed price, settled at the hub
  • A flat block at a fixed price
Answer

The flat block. Every contract here is settled in cash, and the block's settlement fixes the buyer's bill. Under as-generated contracts the settled MWh follow the panels, so the bill still moves.

Numbers, sources and what’s next

Numbers to remember

Number What it means Source
$3.9M to $8.2M The solar farm's yearly sales with no contract, P10 to P90 of 200 simulated price years (seed 2025). This lesson's simulation (illustrative drivers)
$5.3M to $6.7M The same, with a fixed price settled at the hub: output and basis stay with the seller. This lesson's simulation
$7.59M The buyer's bill every year under a flat block at a fixed price. This lesson's simulation
24.0% The capacity factor of U.S. utility-scale solar photovoltaic generators in 2025. EIA, Electric Power Annual, Table 4.08.B · final 2025 data, October 2026

Keep going

Back to the course map

Sources and further reading

Spotted an error, or stuck? Write to hlli@ksu.edu. Corrections are made in place and dated here.

Text licensed CC BY 4.0, code MIT: teachers and students may reuse and adapt it, with credit to Honglin Li.