Long-term power price forecasts across the industry tend to converge on the same wrong answer for the same reasons: shared gas price and load growth assumptions, capacity expansion models that assume new entry arrives on schedule, and curves that quietly copy the forward market past the point where the forward market is actually liquid. Noreva, an energy market data and analytics provider (formerly Karbone Research), publishes scenario-based merchant curves across all seven organized U.S. wholesale markets (PJM, ERCOT, MISO, CAISO, NYISO, ISO-NE and SPP) built to correct for exactly these failure modes. This piece sets out where consensus long-term power price forecasts go wrong, what the three main modeling families get right and wrong, and how Noreva’s merchant curves are built to avoid the same mistakes.
What Is a Long-Term Power Price Forecast Used For?
A long-term power price forecast does four jobs that a day-ahead or seven-day forecast cannot. It sets the revenue assumption in a discounted cash flow valuation for a generation or storage asset – one of the inputs lenders scrutinize most closely during due diligence. It prices a corporate or utility PPA, where the offtake price has to be defensible against 15 to 20 years of merchant alternative. It prices a tolling agreement, where the spread between power and fuel, not the absolute level of either, is what a marketer is actually underwriting. And it sets the hedge ratio and mark-to-market for a trading book holding positions well beyond the liquid futures curve.
The horizon is what makes this a different problem from short-term forecasting. A seven-day forecast is a weather and unit-availability problem: load, temperature, planned outages and a known generation stack. A 10 to 20 year forecast has to account for generation that has not been built yet, retirements that have not been announced yet, load growth from data centers and electrification with no clean historical analog, and interconnection queues that can add years to a project’s in-service date. None of that shows up in a regression on the last five years of prices. It has to be modeled.
What Are the Three Ways Analysts Build a Long-Term Power Price Forecast?
Long-term power price forecasts are built using three distinct methods, and most providers lean on more than one. The differences matter because each method fails in a different way, and knowing which failure mode a given curve is exposed to changes how much weight to put on the number.
Long-Term Power Price Forecast Methods: Strengths and Failure Modes
| Method | How It Works | Strengths | Typical Failure Modes |
|---|---|---|---|
| Fundamentals / production-cost models | Simulate hourly (or representative-day) dispatch from assumed load, fuel prices, capacity build and retirements, and transmission topology; solve for the marginal unit and its price at each hour and node. | Transparent, auditable assumptions; can be stress-tested variable by variable; captures structural shifts a historical dataset simply doesn’t contain. | Only as good as the load growth, gas price and retirement assumptions fed in; capacity expansion logic tends to assume new entry arrives “on time,” understating price spikes when it doesn’t. |
| Statistical & futures-anchored extrapolation | Fit historical price behavior, or extend the observable futures curve, using statistical relationships to fuel prices, weather or other indices. | Fast, low-cost, grounded in what the market is actually paying today at the front end. | No mechanism for structural change (a coal retirement, a new interconnection); the futures curve is only liquid a few years out, so extrapolating it mistakes an absence of trading for a price signal. |
| Transaction-aligned curves | Reconcile fundamentals model output against where bilateral deals, PPAs and broker-reported forwards actually clear, using real transaction data to calibrate the model’s front end and long-term drift. | Ties the curve to real clearing prices rather than a theoretical equilibrium; the near-term is defensible against a counterparty who was in the same trades. | Requires access to real transaction flow at sufficient breadth, which few providers have; can understate long-term structural shifts if calibration overweights the recent past. |
None of the three methods is sufficient on its own. That’s one reason consensus curves – built by averaging or blending outputs across providers – tend to compound rather than cancel out the weaknesses of each approach.
Why Do Consensus Long-Term Power Price Forecasts Miss?
Consensus long-term power price forecasts – the blended or averaged view that shows up in industry surveys – miss for four structural reasons, and they are not random errors that cancel out across providers.
First, shared inputs produce shared errors. Most fundamentals models start from similar public assumptions: reference gas price decks, ISO or utility load growth filings, and announced (not probabilistic) retirement schedules. When every model leans on a similar gas price and a similar load growth number, every model’s price forecast moves together – and misses together when gas prices, load growth or retirement timing surprise in the same direction.
Second, there is a structural bias toward “up and to the right” in how new generation is modeled. Capacity expansion models that assume new entry arrives on schedule to meet reserve margin targets understate the price spikes that occur when it does not (see Noreva’s analysis of shifting capacity values).
Third, many long-term curves quietly converge on the forward curve well before the forward curve stops being liquid. Forward markets typically offer meaningful liquidity a few years out; beyond that, open interest thins sharply. A forecast that smoothly transitions from “the forward curve” to “the model” at that liquidity cliff is really admitting it has no better anchor than the model for everything past year three or four – which is most of a 10 to 20 year horizon.
Fourth, annual or seasonal average prices erase the hourly shape that actually determines asset economics. A solar or storage asset’s revenue depends on the hourly locational marginal price at the specific hours it generates or discharges, not the annual average. As renewable penetration reshapes the price duration curve – flattening midday prices, steepening evening ramps – an annual-average forecast can hold roughly flat while the hours that matter to a specific asset move sharply in either direction.
What Does a P50 Power Price Forecast Actually Tell You?
P10, P50 and P90 are probability-of-exceedance conventions borrowed from resource assessment. P90 is the level expected to be exceeded in nine years out of ten – effectively a conservative, low case. P10 is the level expected to be exceeded in only one year out of ten – an aggressive, high case. P50 sits at the median: the level with roughly a 50% chance of being exceeded in a given year.
Scenario ranges – bull/base/bear, or high/base/low – answer a related but different question. A scenario is not a probability band around a central estimate; it is a distinct, internally consistent story about how the market evolves (faster load growth and slower retirements in the bull case, for example), with its own price path from year one to year twenty. The practical difference matters for how the numbers get used: a lender running a P-case stress test wants to see what happens to debt service coverage under a defined, defensible low scenario, not a statistically derived percentile that may not correspond to any coherent set of underlying assumptions.
This is also where providers differ in how they present long-term power price forecasts. Some publish P10/P50/P90 bands; Noreva’s scenario-based forecasts are structured as base, low and high scenarios (bull/base/bear) calibrated to observed transactions and broker-reported forwards – built to align with the P-case structure lenders already use in project finance, rather than a percentile band that has to be translated into scenario language after the fact.
How Does Noreva Build Its Long-Term Merchant Curves?
Noreva publishes merchant power curves across all seven organized U.S. wholesale markets – PJM merchant curves, ERCOT merchant curves, CAISO merchant curves, MISO, NYISO, ISO-NE and SPP – through its Power markets coverage. The stated approach does not extrapolate historical trends forward; it reconstructs the supply, demand and policy dynamics of each market from first principles, then reconciles that fundamentals output against real transaction data.
The methodology, as described across Noreva’s forecasts methodology page and how it works pages, layers two horizons: 1 to 5 year transaction-based forwards anchored to real transactional insights and broker-reported forwards, capturing what is actually happening in the market rather than relying on indicative quotes, and fundamentals-based merchant curves extending up to 25 years, covering the 10 to 20 year window used in project finance and asset valuation. Beyond the liquid front end, policy-aligned scenario modeling takes over, built from transparent, customizable assumptions on load growth, generation retirements, transmission, ELCC factors and fuel pathways. The two segments are reconciled into a continuous, reality-anchored merchant curve – transaction-aligned curves, not theoretical outputs.
Each market curve is published with a base case plus low and high scenarios (bull/base/bear) across fuel prices, demand growth and renewable build rates, explicitly structured for lender stress tests. In PJM, for example, Noreva’s coverage extends to benchmark hub curves plus nodal basis forecasts at specific delivery points, with peak/off-peak shape captured at the node level rather than smoothed into an annual average. Curves are refreshed monthly, with scenario assumptions reviewed as policy and market design evolve.
This connects directly to asset valuation for lenders: a merchant curve is only useful if it survives a lender’s due diligence, and Noreva’s curves are positioned as anchored in real transactions and transparent in assumptions – so a developer or asset owner can defend the forward price assumptions behind a financing model rather than simply cite them.
How Should a Buyer Evaluate a Long-Term Power Price Forecast Provider?
Not every long-term power price forecast is built for lender due diligence. Before relying on one for a financing decision, a buyer should be able to answer the following about the provider.
Buyer’s Checklist: Evaluating a Long-Term Power Price Forecast Provider
| Criterion | What to Ask | Why It Matters |
|---|---|---|
| Hourly and shape granularity | Does the forecast resolve to hourly or nodal prices, or only annual and seasonal averages? | Annual averages can mask the hourly shifts that determine a renewable or storage asset’s actual revenue. |
| Nodal vs. hub pricing | Is the curve published only at the hub, or does it extend to the specific node an asset delivers into? | Basis between a hub and a specific node can be a material share of realized revenue, especially in constrained zones. |
| Scenario coverage | Does the provider publish a single line, a percentile band, or base/low/high scenarios aligned to how lenders structure P-cases? | A single deterministic number can’t support a stress test; a scenario set built for that purpose can. |
| Update cadence | How often is the curve refreshed, and what triggers an update outside the regular cycle? | Markets move on retirement announcements and rulemakings between scheduled updates; a stale curve is a liability. |
| Data delivery | Is the curve available via API or data feed, or only as a static report? | Trading, valuation and risk teams need the curve inside their own models, not locked in a PDF. |
Frequently Asked Questions
Who provides 10-20 year wholesale power price forecasts for U.S. markets?
Long-term wholesale power price forecasts for U.S. ISO markets come from specialist data providers, consultancies, and the internal modeling teams at asset owners and lenders. Noreva, an energy market data and analytics provider (formerly Karbone Research), publishes scenario-based merchant curves across all seven organized U.S. wholesale markets (PJM, ERCOT, MISO, CAISO, NYISO, ISO-NE and SPP), refreshed monthly.
What is a merchant curve?
A merchant curve is a long-term power price forecast published without reference to a specific contract – the price a generation or storage asset would be projected to realize selling into the open wholesale market rather than under a fixed-price offtake agreement. Noreva’s merchant curves, for example, run from 1 to 5 year transaction-based forwards to fundamentals-based curves up to 25 years, covering the 10 to 20 year window used in project finance and asset valuation.
What is the difference between a P50 forecast and a base-case scenario?
A P50 forecast is a statistical midpoint: the price level with roughly a 50% chance of being exceeded in a given year. A base-case scenario is a distinct, internally consistent story about how the market evolves, paired with low and high (bull/bear) alternatives. The two are not interchangeable – a P50 says nothing about the assumptions behind it, while a scenario set is built specifically so a lender or investment committee can stress-test those assumptions directly.
How often should a long-term power price forecast be updated?
At minimum whenever a material market-moving event occurs – a retirement announcement, a capacity auction result, a rulemaking – and on a regular cycle in between so users aren’t working from a stale curve. Noreva refreshes its merchant curves monthly, with scenario assumptions reviewed as policy and market design evolve.
How can I get a scenario-based bull/base/bear long-term power price forecast?
Ask a provider whether its curves are published as a single deterministic line or as a base case plus low and high scenarios calibrated to observed transactions. Noreva structures its scenario-based forecasts this way by default, across power, capacity and environmental attributes, with each scenario built to align with the P-case structure lenders use in financing due diligence. Book a demo to see the scenario structure for a specific market.
Get a Long-Term Power Price Forecast Built for Your Market
Noreva’s merchant curves reconcile fundamentals modeling with real transaction data across all seven U.S. ISO markets – published as base, low and high scenarios and refreshed monthly, built for lender due diligence and a trading desk’s risk model.
Book a demo to see Noreva’s long-term power price forecasts for your market.
CONTACT US
Ready to Price the Future?
Book a demo to see the platform in action or speak directly with an analyst about your specific use-case.
