1 · Customer avoided cost
The customer saving is usually the largest and easiest figure to model in an early-stage industrial solar and battery case, and it is usually real. They should not be treated as project revenue unless the contract converts them into payments to the project company. A remote site burning diesel really does save a great deal of money by displacing fuel. The mistake is to assume that number belongs to the project company.
Early models often start with customer data such as fuel use, tariffs and load, because that is what is available first. That shows the customer’s potential saving, not the revenue available to repay project debt. A lender asks which contracted payments are available for debt service. The project may still create value, but the model does not yet show how that value will repay debt.
A solar + BESS project displaces diesel, lowers generator O&M, improves reliability and creates carbon benefit for the site owner. Together these are the customer’s avoided costs: the full value the project creates for the energy user. That value belongs to the customer unless the contract transfers part of it.
The project company receives only what the contract says it receives. A lender then sizes debt against the portion of that revenue that is contracted, durable and protected under its downside case. Three different amounts are involved: what the customer saves, what the project company is paid, and what a lender will lend against.
2 · Contracted SPV revenue
The useful early-stage discipline is to follow value across the whole project:
Industrial BESS cash flow chain
The amount narrows at each step. The lender sizes debt against the contracted cash flow that remains after costs and downside adjustments.
The amount narrows at each step. If the fee captures only part of the saving, it narrows at the first step. If the fee is uncontracted or contingent, it narrows at the second. If the term is short or the coverage ratio thin, it narrows at debt service. The lender underwrites the amount of contracted cash flow that remains after costs and its own downside adjustments.
The parties and payments in a behind-the-meter project line up as follows: the party receiving the benefit, the party paying the SPV and the parties carrying technical and delivery risk have to be the same set, or connected by contract.
3 · Lender-recognised cash flow
The analysis separates project revenue into three cases, each answering a different party’s question:
Fixed, contracted EaaS fee only
The predictable payment obligation a lender can actually underwrite. This is the foundation. If S1 does not clear the DSCR threshold, the project is not financeable under the current structure — however good the customer economics look.
Fee plus behind-the-meter and customer-side value
Additional value — arbitrage, spare-capacity use, the customer’s wider avoided cost. It strengthens the equity case and the customer’s willingness to sign. But it depends on dispatch rights and behaviour, so a careful lender discounts or excludes it.
The stressed, bankable case
S1 run through the lender’s pessimism: lower availability, customer under-use, delay, credit stress. This is the case debt is actually sized against. If the project fails under the stressed contracted-revenue case, the model does not yet support the proposed debt structure.
S1 determines debt capacity, S2 adds potential equity value, and S3 tests whether debt remains supportable under downside assumptions.
Illustrative worked example
| Item | Customer | SPV | Lender case |
|---|---|---|---|
| Avoided diesel and tariff cost | A$X | — | — |
| Contracted service fee | (A$Y) | A$Y | A$Y, after the lender’s downside assumptions |
| Uncontracted value (upside sharing, carbon) | A$Z | A$Z | A$0 |
| Operating cost | — | (A$O) | (A$O), stressed |
| Debt service | — | (A$N) | tested at 1.30× DSCR on the contracted fee alone |
Illustrative structure; not derived from a completed transaction. The customer’s saving (X) exceeds the fee (Y) or the contract is not signed; the lender sizes debt on Y net of haircuts, never on X or Z.
An explicit but unfavourable network charge can be modelled, haircut and priced. A material charging rule that remains undefined after 2029 is harder: it may prevent lenders from forming a credible downside case. The same logic applies to connection constraints, marginal loss factors and market redesign.
A generic ±10% regulatory sensitivity is not enough when a rule change can alter whether the asset may participate, when it can energise or which network charges it must carry.
Arbitrage, FCAS, capacity and balancing are often presented as separate columns. In operation, the battery buys and sells across markets while constrained by the same state of charge, availability, efficiency, degradation and connection capacity. Charging may appear as negative revenue in one market before enabling a profitable discharge elsewhere.
A financeable model therefore needs a single physical ledger:
- interval charge and discharge volumes and prices;
- state of charge, efficiency, availability and degradation;
- net settlement across all markets;
- optimiser fees, imbalance and redispatch outcomes;
- conflicts where the same capacity cannot perform two services simultaneously.
Compare these streams by the net revenue each adds per cycle and per MW, after the cost of the cycling they require.
4 · Demand-charge reduction, as one example
For customers on demand-based network tariffs, peak-demand reduction is the value stream that can be attributed most directly to a battery, and the one most easily converted into a contracted fee. A commercial bill combines an energy charge, fixed charges and a demand charge. The energy component measures how much electricity the site uses; the demand component measures how fast it uses electricity at its highest point.
Two facilities can use very different amounts of energy and still face the same demand charge if they reach the same peak. A compressor, chiller, crusher or group of fast chargers can set that peak in a single interval. A battery behind the main meter discharges to hold a target demand ceiling, then recharges when load and tariff conditions allow. It does not need to reduce total consumption.
The control problem is harder than it sounds. The system must decide how much charge to keep in reserve and which peak is likely to matter, because one missed peak can reset the billing maximum and remove most of the month’s expected saving. Australian commercial and industrial tariffs recover part of network cost through measured peak kW or kVA, sometimes within defined time windows and sometimes with ratchet effects, so the opportunity is specific to the site and the tariff. The first task is therefore not choosing a battery. It is reconstructing the delivered cost from bills, interval data and the applicable network tariff.
5 · Other value streams
The same battery may support several value streams:
- Peak-demand shaving against a kW or kVA network charge.
- Tariff optimisation, including staying below a threshold that changes the applicable tariff.
- Energy arbitrage between low- and high-price periods.
- Demand response, FCAS or network support, where market access and operating constraints permit.
- Solar self-consumption by shifting surplus generation into a later load period.
Each benefit in a stack has to be deliverable at the same time as the others. Each service competes for power, energy capacity and state of charge. The defensible case is the compatible stack the control system can actually deliver under the customer’s tariff and operating profile.
Forecast-to-financeability bridge
The forecast is only the first input. Debt capacity appears after physical dispatch and lender downside tests.
Harmony Energy’s experience with two-hour batteries provides a useful way to think about duration. A shorter-duration asset may appear superior while high-value ancillary services are deep and unsaturated. If those services compress earlier than expected, the asset can lose both revenue and strategic flexibility.
Additional duration may lower the central-case IRR while expanding the set of markets the battery can enter and reducing dependence on one service. The investment decision should therefore compare more than the expected return. It should test capital survival under service saturation, spread compression and dispatch error.
For customers under emissions reporting, safeguard obligations or internal decarbonisation targets, diesel displacement carries genuine carbon value. It should be quantified and shown — it is often why the customer is at the table. But it needs discipline: the carbon benefit belongs to the customer unless the contract explicitly assigns it to the SPV, and a lender will not underwrite carbon revenue unless it is contracted, certified and stable, which it rarely is early on. Counting it in both places would double-count the benefit.
6 · Stress case
Failure to meet the lender case may indicate a problem with revenue allocation, contract tenor, credit support or gearing rather than with the underlying technology. The problem may lie in the contract or the financing structure rather than in the battery. The changes that usually matter are:
- a higher fixed fee, so the project captures more of the customer saving;
- a longer contract term;
- a take-or-pay floor under the payment;
- stronger customer credit or termination provisions;
- lower gearing;
- a different repayment profile, where tail-year sculpting has dropped the minimum coverage ratio.
The appropriate conclusion is that the financing structure needs work, not that the project is unviable.
- Size to the shape and volatility of the load. Nameplate peak alone does not reveal how long the event lasts or how predictable it is.
- Derate forecast savings. Volatile loads, solar variability, outages and changing operating schedules reduce perfect-model performance.
- Check tariff rules. Billing windows, kVA measurement, ratchets and threshold changes can matter more than the nominal demand rate.
- Separate resilience. A battery sized to clip short peaks is not automatically capable of carrying critical load through an outage.
- Test degradation and availability. The valuable peak may occur when the battery is unavailable, energy-limited or preserving warranty constraints.
- Debt foundation: repeatable or contracted revenue with transparent lender haircuts.
- Operating scenarios: service saturation, spread compression, availability, degradation, delay and optimiser underperformance.
- Equity optionality: upside from volatility, new services and rule changes without relying on it to support DSCR.
COD delay should also flow through equipment repricing, contract commencement, interest during construction and the revenue curve. Simply shifting revenue one year to the right understates the damage.
The customer does not always need to fund the asset directly. Shared-savings agreements, fixed service payments, solar-plus-storage PPAs and performance contracts can convert an uncertain capital decision into an operating-cost proposition. But the contract must say who carries performance risk when the peak is missed, the tariff changes or the operating profile moves.
7 · Questions to test
Before a system is quoted, the following can be answered from the customer’s interval data, tariff and draft contract. If any answer is missing, the model is describing a value pool, not a financeable project.
Questions to ask
- Which party owns each saving and environmental benefit?
- What payment is fixed, contracted and enforceable?
- How long does the contract run relative to the debt and asset life?
- What remains after availability, operating cost and lender haircuts?
- Does the stressed CFADS still cover debt service?
- For a demand tariff, which interval sets the charge? How predictable is the peak, and how much can the proposed battery reduce on the worst day?
- Which other services can the same battery deliver at the same time as peak defence, under the customer’s operating profile?