Generation Forecast Models
Executive Summary
Key Takeaways
- ✓ A generation forecast is distinct from a resource yield assessment — the yield assessment establishes confidence-level output figures (P50, P90), while the forecast builds the full time-series schedule (monthly or finer) the financial model actually runs its revenue and debt sculpting calculations against.
- ✓ Forecast granularity should match the frequency at which the model tests revenue and coverage, since an annual-only forecast can conceal sub-annual periods of low output that create genuine covenant stress.
- ✓ An explicit uncertainty band, not just a single central forecast line, should be carried into sensitivity and downside testing, since the forecast's confidence level materially affects how much revenue risk the model actually represents.
- ✓ Forecasts should be updated periodically against actual operating data once the asset is operational, reconciling forecast-to-actual variance rather than holding the original pre-construction forecast fixed indefinitely.
- ✓ Generation forecasting methodology should be technology-appropriate — combining resource data with the specific technical output chain (irradiance-to-AC for solar, wind speed-to-power-curve for wind, and so on) rather than a generic output growth assumption.
Objective¶
This guide covers how to build a generation forecast within Energy Financial Modelling, distinguishing it from the resource yield assessment it draws on and setting out the granularity, uncertainty, and update practices a financial model should apply.
Resource Yield Assessment vs. Generation Forecast¶
A resource yield assessment establishes the underlying resource data at defined confidence levels — P50 and P90 exceedance figures for solar irradiance, wind speed, or hydrology. A generation forecast is a distinct, subsequent exercise: building the full time-series output schedule the financial model actually runs its revenue, cost, and debt sculpting calculations against, by combining that resource data with the project's specific technical output chain — panel and inverter specification for solar, turbine power curve for wind, and so on. Treating the yield assessment's single confidence-level figures as if they were themselves a ready-to-use forecast skips this translation step.
Forecast Granularity¶
The forecast should be built at a granularity matching the frequency at which the model tests revenue and debt service coverage — typically monthly at minimum for assets with material seasonal output variability. An annual-only forecast can conceal sub-annual periods of low output (a low-wind season, a dry-season hydrology period) that create genuine coverage ratio stress even where the annual total appears entirely adequate, particularly where actual covenant testing itself occurs at a sub-annual frequency.
Carrying Forward an Explicit Uncertainty Band¶
The forecast's central line should be accompanied by an explicit uncertainty band reflecting the underlying resource assessment's confidence level, carried through into sensitivity and downside testing. Presenting only a single central forecast figure implies a precision the underlying resource data does not actually support, and a lender or investment committee needs to see the range, not just the point estimate, to properly assess revenue risk.
Updating the Forecast Against Actual Performance¶
Once the asset is operational, the forecast should be periodically reconciled against actual generation data, updating forward assumptions where actual performance diverges materially from the original pre-construction forecast. Holding the original forecast fixed indefinitely, regardless of demonstrated actual performance, forgoes the most valuable evidence available once the asset has an operating track record — consistent with the treatment described in Infrastructure and Energy Transactions for secondary-market transactions relying on actual operating history.
Common Construction Pitfalls¶
Yield assessment treated as a ready-made forecast. Using the resource yield assessment's single confidence-level figures directly as the model's output schedule, without building the full technical conversion chain, skips a necessary modelling step.
Annual-only granularity. Building only an annual forecast when covenant or revenue testing occurs sub-annually can conceal genuine periods of coverage stress.
No uncertainty band carried forward. Presenting a single central forecast without its associated confidence range understates genuine revenue risk in sensitivity testing.
Recommended Practices¶
- Build the forecast as a distinct step from the resource yield assessment, applying the technology's specific technical output chain.
- Match forecast granularity to the frequency of revenue and covenant testing in the model.
- Carry the resource assessment's confidence-level uncertainty band into sensitivity and downside testing.
- Reconcile the forecast against actual operating data periodically once the asset is operational.
Continue Reading¶
Related Pillars¶
Related Technical Guides¶
Related Glossary¶
How OXXON tests thisRun a free structural check with FMAE
Frequently Asked Questions
How does a generation forecast differ from a resource yield assessment?
The resource yield assessment establishes the underlying resource data at defined confidence levels (P50, P90); the generation forecast is the full time-series schedule built from that data, combined with the specific technical output chain (panel and inverter specification, turbine power curve, and so on) that the financial model actually runs its calculations against.
What granularity should a generation forecast use?
Granularity matching the frequency at which the model tests revenue and debt service coverage — typically monthly at minimum for assets with material seasonal variability, since an annual-only forecast can conceal sub-annual periods of low output that create genuine covenant testing stress even where the annual total appears adequate.
Why does the forecast need an explicit uncertainty band?
Because the central forecast line alone does not communicate how confident that estimate actually is — carrying the uncertainty band (reflecting the underlying resource assessment's confidence level) into sensitivity and downside testing shows how much revenue risk the model actually represents, rather than presenting a single figure with implied but undisclosed precision.
Should the forecast be updated once the asset is operational?
Yes — the forecast should be periodically reconciled against actual operating data once available, updating forward assumptions where actual performance diverges materially from the original pre-construction forecast, rather than holding the original forecast fixed indefinitely regardless of demonstrated actual performance.
Does forecasting methodology differ by technology?
Yes — the forecast should combine the relevant resource data with the specific technical output chain for the technology in question (irradiance-to-AC conversion for solar, wind speed-to-power-curve conversion for wind, flow duration curve conversion for hydro), rather than applying a generic output growth assumption that ignores the technology-specific conversion mechanics.
References
Related Articles
Energy Financial Modelling
Energy financial modelling is the discipline of building financial models for power generation assets, independent power producers, and renewable energy projects — structured around a technical output schedule and an electricity revenue stack that a standard corporate or general project finance model has no direct equivalent for. This page is the hub for the Knowledge Centre's energy and power modelling content: how a power project model is architected, how electricity markets and dispatch mechanics translate into revenue, and how power purchase agreements, capacity payments, and merchant exposure combine into a project's revenue structure. Technology-specific renewable energy models (solar, wind, storage, hydro, and others), technical and commercial modelling mechanics, and institutional practice for this asset class are indexed here as the domain expands.
Resource Yield Assessment
A resource yield assessment is a technical study, typically prepared by an independent engineer, estimating the expected energy resource available to a generation asset — solar irradiance, wind speed, or hydrology — expressed at defined confidence (exceedance probability) levels such as P50 and P90. Each confidence level serves a distinct modelling purpose, and using the wrong one for a given purpose is a common structural error in renewable energy financial models.
Power Project Financial Model Structure
A power generation financial model is architected around a technical output schedule — generation volume for a variable-output asset or available capacity for a dispatchable one — that drives every downstream calculation: the electricity revenue stack, the operating cost build, and, where the asset is project-financed, debt sculpting and covenant testing. This guide sets out that architecture as a sequence of explicit, separately built modules, distinct from a standard corporate model's revenue-growth-first structure.
Degradation Modelling
Degradation modelling methodology goes beyond applying a flat annual percentage: it involves choosing between a linear and non-linear degradation curve shape appropriate to the technology, sourcing technology-specific degradation profiles, reconciling a warranty-guaranteed rate against actual measured performance once operational, and understanding how degradation feeds refinancing and repowering decisions later in the asset's life. This guide covers the methodology behind the degradation schedule already introduced as a core technical mechanic.