With abundance of work stream on project finance deal, we sometimes overlook the importance of financial model. Financial model, probably one of the most important elements of the deal which bring together all assumptions and convert them in the investment metrics which sponsors are looking for.
There is a version of a BESS financial model that gets a developer to investment committee. And there is a version that gets a project to financial close. They are not the same model.
The gap between them tends to surface around week three of due diligence, when a Technical Adviser returns a query list. By that point, the financing timeline has slipped, the lender’s credit committee slot has moved, and the developer is repricing assumptions under time pressure. Most of that pain is avoidable — if the model was built to survive lender scrutiny rather than to support an internal case.
What follows is a direct account of the five areas where BESS models most consistently break down under lender review:
- capture rate,
- degradation,
- dispatch efficiency,
- revenue correlation, and
- reserve account stress
— are the standard due diligence checklist for any experienced infrastructure debt team reviewing a UK BESS project in 2026.
1. Capture rate — and the data vintage behind it
The first thing a lender’s model team does when they open a project model is find the revenue assumption and trace it to its source.
A capture rate is the percentage of the reference market price a project actually realises in dispatch. For a merchant BESS project relying on wholesale arbitrage, a modelled capture rate of 85–90% is defensible with good data. Above 92%, questions start. Not because the number is impossible, but because the data required to justify it is rarely as robust as the model implies.
The problem compounds quickly. Suppose a project’s modelled revenue is £45/MWh and the capture rate drops from 92% to 85% in a lender’s independent assessment. Annual revenue falls by roughly 7–8%. In a project sized to a 1.35x DSCR, that movement can push the lender’s downside case to 1.08x — below the floor at which most infrastructure debt funds will hold their current structure without repricing or requesting additional credit support.
What lenders want to see is a capture rate derived from observed dispatch performance on comparable operational assets, not a theoretical optimisation. Modo Energy’s operational data — disaggregated by technology, duration, and market participation — is now granular enough to do this properly for UK BESS. A rate built from that data, with the methodology shown, will survive a TA review. A rate sourced from a consultant’s optimisation model, without an operational comparable, usually will not.
One additional pressure point in 2026: the DESNZ/Ofgem joint letter on connections reform published this month flags a material surplus of BESS capacity progressing to Gate 2 relative to the Clean Power 2030 Action Plan ranges. Lenders are beginning to apply a market saturation haircut to long-run merchant revenue assumptions — particularly for projects commissioning post-2027. If your capture rate assumption holds flat over a 15-year horizon without addressing this dynamic, expect the question.
2. Degradation — the assumption that is almost always wrong in the later years
A BESS project models a 15–20 year asset life. Degradation — the decline in usable capacity over time — is empirically grounded in years 1–5, where operational data exists. In years 8–15, it is largely theoretical, and the theoretical assumptions in most models are optimistic in ways that matter to lenders.
The standard approach is a linear degradation curve, typically sourced from the manufacturer warranty. This creates two compounding problems. First, manufacturer warranties define minimum guaranteed performance — they are a contractual floor, not a performance forecast. A model that uses the warranty curve as its base case is effectively assuming the asset performs at its best contractually-guaranteed level throughout its life, which overstates expected throughput to what a prudent analyst would use. Second, real-world battery degradation is not linear. It tends to steepen in later years as thermal cycling accumulates, state-of-charge management becomes more constrained, and cell chemistry moves further down the capacity curve.
A competent TA will run an independent degradation model — typically calibrated to published academic literature and operational fleet data — and compare it to the sponsor’s. If the sponsor’s model shows 82% of nameplate capacity in year 12 and the TA’s shows 74%, the 8-percentage-point gap is not academic. On a 100 MW project at 250 cycling-equivalent days per year, the revenue difference over a 3-year period in the middle of the debt tenor is £6–10m. That reopens the debt sizing conversation.
What lenders want to see is a degradation curve that is explicitly conservative relative to the warranty floor, with the methodology stated, and an augmentation plan that is specific: when capacity is added, at what cost, who supplies it, and how that cost is funded. If augmentation capex sits in a reserve account that also covers O&M shortfalls, the model should show what happens when both draw simultaneously — because that is precisely the scenario a lender’s credit committee will construct.
3. Dispatch logic — the gap between modelled and realised
Dispatch is where the largest single gap between modelled and actual BESS performance lives, and it is the hardest assumption for a lender to validate independently.
Some models contain a dispatch algorithm — a set of rules governing how the asset moves between revenue streams across a given settlement period. In the model, the algorithm is frictionless. It charges when prices are low, dispatches into the Balancing Mechanism at the optimal price, captures the FFR window in full, and never misses a tripling opportunity. In an operating asset, dispatch optimisers have latency. Battery management systems impose state-of-charge constraints that override the commercial logic. The market moves in ways the algorithm was not trained on.
The gap between modelled and realised dispatch — sometimes called dispatch efficiency — is rarely built into a sponsor’s base case. Most models assume the revenue stack is captured at close to 100% efficiency and leave the gap to be discovered by the TA. A lender’s TA will typically apply a haircut of 5–15% to modelled revenues to reflect this friction. At the lower end, for a project with an established operational track record and a well-documented optimisation system, 5% is defensible. For a project with no operational comparable and a bespoke dispatch system, 15% is not uncommon.
If your model does not already contain a dispatch efficiency assumption, it will be added by someone else — and their number will be more conservative than yours. The better practice is to include it explicitly, explain the basis for the chosen percentage, and reference operational data from a comparable project if one exists. A model that applies its own 8% efficiency haircut and shows the methodology is materially more fundable than one that implies 100% efficiency and waits for the TA to cut it.
4. Revenue stack correlation — the stress that most models do not run
BESS projects typically model revenues from several streams: Dynamic Frequency Response, the Balancing Mechanism, wholesale arbitrage, and the Capacity Market. The model constructs a base case by summing the expected contribution of each stream. The lender’s credit committee deconstructs that base case by asking what the portfolio looks like when multiple streams compress simultaneously.
This is not a hypothetical stress. FFR volumes in GB have been contracting as system inertia improves and more storage capacity enters the market. BM revenues compress as more assets compete for the same dispatch windows at similar marginal costs. These are not independent risks. The market environment that reduces FFR volumes — more storage on the system — is typically the same environment that tightens BM spreads. A model that treats each revenue stream as independently stable in its downside case is not capturing the correlation.
The consequence is that lender-constructed downside cases tend to be more severe than sponsor-constructed downside cases for the same project, because the lender runs the correlated scenario and the sponsor does not.
In a credit committee, an unexplained gap between the sponsor’s downside DSCR and the lender’s own calculation creates negotiating friction that delays the close. What lenders want to see is a revenue stack that shows each stream’s percentage contribution to total revenue, a stream-level bear case for each, and — explicitly — a portfolio downside that correlates FFR and BM compression.
That scenario does not have to be the base case or the bank case. It has to exist in the model, with a labelled tab, so the lender’s team does not have to construct it themselves. A model that has already run the scenario a lender will run is a model that accelerates due diligence.
5. Reserve account stack — the simultaneous draw scenario
Reserve accounts are where lenders price their residual anxiety about a project. A Debt Service Reserve Account sized to 6 months of debt service is standard. A Major Maintenance Reserve Account covering the next scheduled augmentation event is reasonable. An operating cost reserve covering 3 months of fixed costs is defensible.
What most models do not stress-test is the simultaneous draw — the scenario in which a revenue shortfall triggers a draw on the DSRA, an augmentation event is due and draws on the MMRA, and an unplanned O&M event draws on the operating reserve in the same quarter. Individually, each reserve account looks correctly sized. In aggregate, the total liquidity requirement in a simultaneous draw scenario can reach 18–25% of annual project revenue. That is a real constraint on distributions and, in a thin-margin project, a real question about whether the reserve accounts can be replenished from operating cash flow without triggering a cash trap.
This scenario does not occur frequently. But it occurs in exactly the conditions — revenue stress plus an unplanned technical event — that also represent the period of maximum lender concern. Credit committees run it. If the model does not address it, the question comes back from the committee.
What lenders want to see is a waterfall that explicitly models the simultaneous draw, shows the replenishment timeline, and identifies the resolution mechanism — sponsor support, a revolving credit facility, or prioritised draws by account type. A model that has stress-tested this and shows a clear answer is not creating a new problem. It is closing one before it is raised.
What this means for model preparation
These five areas — capture rate, degradation, dispatch efficiency, revenue correlation, and reserve account stress — are the standard due diligence checklist for any experienced infrastructure debt team reviewing a UK BESS project in 2026.
The practical implication is straightforward. Before approaching lenders, run the model that a lender’s TA will run: apply a conservative capture rate supported by operational comparables, use a degradation curve that is explicitly below the manufacturer warranty, include a dispatch efficiency haircut with a stated basis, run the correlated revenue downside, and model the simultaneous reserve draw. These action will help you to understand where the DSCR lands.
If it stays above 1.20x in the downside, the project is fundable and the financing process will be orderly. If it drops below 1.10x, the project has a structural issue that is better resolved before mandate than during due diligence — and that is a better outcome for everyone at the table.
Advance Klimate Solutions advises developers, PE-backed platforms, and institutional lenders on renewable energy project finance, financial modelling, and portfolio-company finance across the UK, Europe, and CEE.
www.advaklim.com
Leave a Reply