Renewable Energy Project Finance Modeling

Final project for INAF U6326 — Renewable Energy Project Finance Modeling at Columbia SIPA, Spring 2025. A full-cycle financial model for a hypothetical 200 MW onshore wind farm in West Texas, built from the lender's seat: construction loan through conversion to term loan, debt sculpting, tax-equity flip, P-case stress tests, and a term-sheet draft for the EPC agreement.

What the model covers

• Twenty-five years of operating cash flows driven by a 15-year PPA with a creditworthy offtaker, layered onto a 60/40 debt-equity stack
• Tax equity via PTC and ITC structures, modeled side-by-side so the sponsor sees the trade-off in IRR, leverage capacity, and flip timing — the PTC case ran a longer flip but produced higher absolute equity returns; the ITC case unlocked more leverage up front but capped sponsor IRR after year nine
• Debt sculpting against a 1.30× minimum DSCR (rather than flat amortization) so principal payments move with the revenue curve. The sculpt was solved iteratively in Excel — Goal Seek on each year's debt service against the target DSCR, then back-checked against the LLCR
• P50 / P90 / P99 production scenarios, plus sensitivities on PPA floor, resource year, basis risk, and cost of replacement debt
• Term sheet for the EPC contract, summarized to four pages so a decision-maker who hasn't read the underlying agreement can still see the commercial essence — liquidated damages, performance guarantees, schedule risk, who absorbs the merchant tail

What I built first, and why I rebuilt it

The first version of the model was the obvious one: revenue, costs, depreciation, debt service, equity, done. It told me almost nothing because every line was independent — change the resource assumption and only the revenue moved.

Real project models are structural. Revenue drives DSCR; DSCR drives the sculpt; the sculpt drives leverage capacity; leverage capacity feeds back into the equity IRR. I rebuilt it with those dependencies wired through, and suddenly the model stopped giving me numbers and started giving me arguments. Push the PPA floor down five dollars and watch four downstream sheets disagree about whether the deal still pencils.

That feedback loop is the entire point of project finance modeling. Without it the spreadsheet is just a calculator with a fancy tab list.

The moment I remember

Third week of the assignment, I tightened the PPA floor by $3/MWh — a small change, well inside any negotiation envelope. The DSCR drifted south by 0.04× every iteration through the sculpt. After three loops the deal failed coverage. No amount of tax-equity restructuring was going to out-argue a bad resource year.

That's the lesson — not "tax equity matters" or "PPAs matter," but: the part of the deal you cannot restructure is the part that decides it. Resource years are exogenous. The whole job, on the lender's side, is to figure out what fraction of your collateral is exogenous and price the spread accordingly.

Caveat

This was a course project. The assumptions are calibrated to public NREL benchmarks and a couple of redacted comparable deals; nothing here is investment advice and the offtaker, EPC contractor, and tax-equity investor are stylized, not real. The framework is portable; the specific numbers should not leave the seminar room.

What it changed

I read project-finance term sheets differently now. The legal language is the model's input file. First-loss positions, cash-sweep triggers, lookback ratios — these aren't legal flourishes, they are deterministic expressions of who bears which slice of the production tail. Being able to walk between the spreadsheet and the contract — and to argue in either direction — turned out to be the most useful thing the course taught me.

Course: INAF U6326, Prof. Vivek Mital. Final grade: A.