Underwriting a storage revenue stack without over-trusting the model

A battery model will happily give you a levelized cost and an internal rate of return to four decimals. Almost none of that precision survives contact with a dispatch algorithm, a market rulebook, and a five-year weather record.

That does not mean the model is useless. It means the model is an argument about the future, and the job is to test the argument, not to admire the arithmetic.

The stack is a set of assumptions

Most storage revenue stacks combine some mix of energy arbitrage, capacity, ancillary services, demand response, and, increasingly, bilateral or retail contracts. Each line has its own uncertainty, and they are not equally knowable.

• Energy arbitrage depends on future price spreads, which depend on the same fleet of batteries that is being built.
• Capacity and ancillary depend on market rules that can and do change, and on the asset being available when called.
• Demand response and retail programs depend on counterparties whose terms are rarely locked for the full asset life.

Aggregation into a virtual power plant adds another layer: the value is now a function of how a fleet of assets are dispatched together, how they are compensated, and who bears performance risk when one underperforms.

What to stress, and how

The useful version of a storage review is a set of scenarios, not a point estimate. I focus on the assumptions that move the answer the most:

• Dispatch logic. Is the model running an ideal foresight dispatch, or something closer to what a real algorithm will do with real information?
• Degradation and augmentation. How is capacity fade modeled, and when does augmentation spending land relative to the revenue it protects?
• Round-trip efficiency and availability. Are the efficiency and uptime figures consistent with the equipment and the operating strategy?
• Price formation. For arbitrage-heavy projects, does the spread assumption survive a more competitive future fleet?
• Contract reality. Are program terms modeled as they will actually be paid, including performance penalties and settlement rules?

A model that holds up across a defensible range of these inputs is bankable. A model that only works at its base case is a brochure.

The utility partnership question

Many storage projects only clear because of a utility program, a capacity contract, or a coordinated dispatch arrangement. Those relationships are negotiated, not assumed. The structuring work — how the asset is enrolled, who controls dispatch, how revenue is shared, and what happens if the program changes — usually matters more to the outcome than a tenth of a percent on round-trip efficiency.

The point

Battery diligence is not about catching a math error. It is about separating the parts of the stack that are contracted from the parts that are forecast, and making sure the financing case does not lean on the second as if it were the first. When the assumptions are laid out honestly, the model becomes a decision tool instead of a sales document.