Technical bankability: what the bank and EI require from the resource assessment report

Project finance non-recourse means that the bank lends against the project's cash flows — not against the developer's balance sheet. The park needs to pay. This completely changes what the bank needs to know: it doesn't matter if the developer has other assets; it matters if It's... Project will generate enough energy to serve the debt every year, even in conservative scenarios.

The document that connects the physics of the wind or sun to the financial model of the bank is the Energy Yield Assessment The bank does not read this document directly: it asks that a Independent Engineer Read it, review it, and sign an opinion that the methodology is solid and the numbers are defensible. The technical bankability of the project starts and ends with what the EI finds in the report.

The structure of the renewable finance project

In a typical wind or solar park financing in Brazil, the structure involves:

• Senior lenders (BNDES, commercial banks and DFIs): covered by the AEP P90 in the cash-flow model. They require DSCR ≥ 1.15–1.25 in the base case (P50) and DSCR ≥ 1.05 in the stress case (P90).
• Equity sponsorsThey use P50 from AEP in the analysis of expected return.
• Independent EngineerIt reviews the technical report, issues bankability opinion during the financing process, and monitors construction and operation.

The DSCR (Debt Service Coverage Ratio) is the central coefficient of bank analysis:

DSCR = NOI / Debt_service

NOI = Gross_revenue × (1 − O&M_losses) − OpEx
Gross_revenue = AEP × PPA_tariff

# Typical funding agreements:
DSCR ≥ 1,25  →  base case (P50): no restriction
DSCR ≥ 1,15  →  lock-up: no dividend distribution below this level
DSCR ≥ 1,05  →  event of default: financing in default below this level

Any error in P50 from AEP spreads linearly to the DSCR. A 5% overestimated generation P50 can transform a 1.30 (financial) DSCR into 1.23 (even OK) — or 1.20 to 1.14 (below the lock-up chord). The accuracy of the EYA is not abstract: it determines the leverage limit that the project supports.

The technical chain until closing

The role of P90 in the model of the bank

The bank uses two generation scenarios in the financial model. Basic case It uses P50 — the central estimate, half of the years above, half below. Stress scenario It uses P90 — the production that is only down in 1 in 10 years.

Which P90 to use is a choice that has a technical and financial consequence. P90 at 1 year (σ_total with 1 year u_iav) is the most conservative — it represents the worst typical year. P90 at 10 or 20 years (u_iav diversified in √N) represents the average of a long period in a conservative scenario. Lenders tend to use:

The long-term P90 (10 or 20 years) is the most common in the base model of the legends for projects with 20-year PPP. The 1-year P90 is used in annual covenant clauses — "if the DSCR of any year falls below X, it enters into default event". The two scenarios need to be calculated separately, because total uncertainty is different (AVI does not diversify in the annual test).

The IE checklist: what is under trial

The Independent Engineer does not restate the calculations of the EYA. He checks whether the methodology used is acceptable, whether the input data are defensible, and whether the results presented are consistent with what the methodology should produce. Typical verifications of an EI in the EYA of a renewable park in Brazil:

Where the reports lose bankability

The experience with projects in the Brazilian Northeast points to six recurrent situations of failure or conditioning by EI:

1. Short campaign and uncalculated 1-year P90. 4–6-month campaigns are used to anticipate development, but the EI does not accept them for final EYA without explicit adjustment for the minimum period IEC. The 1-year P90 based on 6-month campaign has substantially higher adjustment uncertainty (u_adjust = RMSE% / √6 instead of √12) and needs to be declared this way.

2. Short-term long-term reference. Use a local weather station with 5 years of history for long-term adjustment is not bankable. The reference must have coverage ≥ 10 years to capture long-term climate cycles (ENSO, AMO, PDO). For the Northeast, ERA5 (since 1940) and MERRA-2 (since 1980) meet this criterion; the choice of reference is discussed in NT.02.

3. Uncertainty budget without decomposition. To declare only "total uncertainty: 8%" without breaking into components (calibration, soiling, adjustment, VAI) does not allow the EI to verify whether the individual values are reasonable. IEC 61400-12-1 (wind) and IEC 61724-1 (solar) require explicit decomposition of the uncertainty components — the budget has to be reproducible by an external auditor, as discussed in NT.06.

4. P90 calculated without separating 1 year vs. long term. The bank needs both: the annual covenant uses 1-year P90 (with u_iav_1yr intact); the base model uses 20-year P90 (with u_iav diversified in √20). Presenting only one value without specifying the horizon is an immediate conditioner of the EI.

5. Data without audit trail. The EI needs to be able to reproduce the EYA calculations from raw data. If the report does not contain hash data files, version of the analysis software, model parameters and processing date, it is not auditable. Technical traceability is discussed in NT.05.

6. MCP correlation with low R² accepted without justification. A MCP with R² = 0.60 may have been calculated with the wrong reference. If the IE see R² below 0.70 without explicit justification (complex terrain, coastal effects, limitation of reanalysis), it conditions: either changes the reference, or documents the limitation with sensitivity analysis of P90 to the adjustment error.

The report as a financing piece

A bankable technical report has to be written with two readers in mind: the engineer who will review it line-to-line (IE), and the credit committee who will see an executive summary of 2 pages. The EYA needs to produce the result table that goes to the financial model:

Result                          P50        P90 (1yr)    P90 (20yr)
──────────────────────────────────────────────────────────────────
Net AEP (MWh/year)              243.800    217.500      235.600
Capacity factor (%)              27,8       24,8         26,9
Total uncertainty σ (%)                 —          5,8%         3,4%
Estimated DSCR (informative)          1,38       1,08         1,22
──────────────────────────────────────────────────────────────────
LT reference: ERA5 (1940–2025), overlap: 14 months, R²: 0,84
Transposition model: Hay-Davies (IEC 61724-1)

This table needs to appear in the executive summary, referenced to the methodological sections where the calculations are detailed. The ABNT report generated by BORA (wind) and SOLANO (solar) structure exactly this format — with numbered sections, reproducible methodology, and hash input data included in the file metadata.

For portfolios with multiple assets, the bank's question changes: are assets correlated? A portfolio with 10 parks in the same wind corridor has less diversification than 10 parks in different states. The aggregation of uncertainties in the portfolio — and how the technical maturity score reflects the bankability of each asset individually before consolidating — is the theme addressed in NT.11.