The Constraint Stack: What Actually Limits the Battery You Can Build
Most battery projects are sized the wrong way round. A developer picks a configuration — 100 MW, four hours, because that is what everyone builds — models the revenue, and then discovers which constraint it has run into.
The reverse order is better, and it is not a stylistic preference. Constraints in a storage project bind on different variables, which means finding the binding one first tells you which of the remaining decisions matter at all.
The point of interconnection limits power. Land limits energy. A capital budget limits both, and in a ratio you choose. So a project can be simultaneously constrained on power and on energy, with neither trading off against the other — and no amount of optimising within one will relieve the other.
This post opens the sizing and dispatch series with the stack itself: what limits a battery, which variable each item binds, and how to find the one that actually decides your project.
ℹ️ Note: Figures are labelled assumptions except where attributed. Fire code and siting requirements are jurisdiction-specific — NFPA 855 is widely adopted but the authority having jurisdiction governs.
What Is the Constraint Stack?
The ordered set of physical, regulatory, commercial and financial limits on the battery you can actually build at a given site. Each binds a specific variable, and the project is determined by whichever binds first — not by the configuration you wanted.
The Seven Constraints
| Constraint | Binds | Typically discovered |
|---|---|---|
| Interconnection capacity | Power (MW) | Cluster study — late |
| Land and footprint | Energy (MWh) | Site control — early |
| Fire code and setbacks | Energy (MWh) | Permitting — middle |
| Capital budget | Both, in a chosen ratio | Investment committee |
| Grid code and technical | Power, sometimes energy | Interconnection agreement |
| Offtake and market | What you can sell, not what you can build | Commercial — variable |
| Supply chain | Delivery date more than size | Procurement — late |
The column that matters is the middle one. A developer that treats all seven as a single "feasibility" question misses that relieving a power constraint does nothing for an energy constraint, and vice versa.
Interconnection: The Filter That Kills Most Sites
By a wide margin the most consequential, and the one that eliminates candidate sites earliest in a screening exercise.
The blunt version, from siting practice: substation proximity is the filter that eliminates the most candidate sites, ahead of acreage, terrain or zoning. A parcel that clears everything else and sits two miles from the nearest usable node is very difficult to make work.
Two features make interconnection distinctive as a constraint.
It binds power, not energy. A 100 MW point of interconnection caps the rate at which the battery can import or export. It says nothing about how many megawatt-hours sit behind that connection — a 100 MW / 200 MWh and a 100 MW / 800 MWh battery present identically to the grid.
That single fact is why duration is the cheap dimension to expand on a constrained site. Once the interconnection is secured, adding hours does not require going back to the queue.
It resolves late and exogenously. As the Series C interconnection posts established, the cluster study assigning network upgrade costs arrives years into development, after most spend is committed, and it moves with other parties' decisions. So the constraint most likely to determine the project is also the one you learn about last.
There is an important asymmetry in how a battery is treated. A storage project can often accept a limited or conditional interconnection service — agreeing to curtail under defined conditions in exchange for a faster or cheaper connection. That is a genuine option for a battery in a way it rarely is for generation, because a battery can shift its operation rather than lose it.
Land: The Arithmetic
Simpler, earlier, and it binds the other variable.
A rough industry benchmark is about 1,000 square feet of footprint per MWh of storage capacity for the containers themselves. A 100 MWh project needs roughly 2.3 acres for containers, with total site area — transformer yard, access roads, setback buffers, perimeter fencing — typically running 5 to 20 acres depending on design and local codes.
Run that against the worked project this series uses:
100 MW × 4 hours = 400 MWh
Container footprint 400 × 1,000 sq ft = 400,000 sq ft
= 9.2 acres
Total site, at the same ratio as a 100 MWh project = 20–80 acres
The multiplier from container area to total site is the part that surprises people. Roughly half to three quarters of a battery site is not battery — it is spacing, access and buffer, and most of that is driven by fire code rather than by engineering preference.
Fire Code: The Constraint That Has Been Tightening
NFPA 855 governs the installation of stationary energy storage systems and is the standard most jurisdictions reference.
The 2026 edition, released in September 2025, made two changes that matter for siting.
Hazard Mitigation Analysis is now mandatory for virtually all battery installations over 1 kWh, broadening what was previously a narrower requirement.
Explosion control is required in accordance with NFPA 69, or a performance-based alternative supported by installation-level fire and explosion testing and engineering evaluation.
On spacing, Section 15.5 requires individual ESS units to be separated by a minimum of three feet, unless smaller separation distances are documented as adequate and approved by the authority having jurisdiction on the basis of large-scale fire testing.
That last clause is where real land is won or lost. A vendor with large-scale fire test results supporting tighter spacing can fit materially more energy on the same parcel than one without — which makes fire testing evidence a siting asset, not merely a compliance document. On a land-constrained site it is worth diligencing before the equipment is selected.
Why the Binding Constraint Has to Be Found First
Because it determines which optimisation is worth doing.
Consider the same project under two different binding constraints.
Power-constrained (a 50 MW interconnection on a large parcel). The MW is fixed and cannot be bought. Every remaining decision is about what to do with those 50 MW — how many hours behind them, which products to sell, how hard to cycle. Duration analysis is the whole game, and it is cheap, because adding hours does not touch the constraint.
Energy-constrained (a 100 MW interconnection on a small urban parcel). The MWh is fixed by land and fire code. Duration analysis is nearly pointless — you cannot add hours. The decisions that matter are about power rating and product mix: with limited energy, high-power short-duration products like frequency response become relatively more attractive.
Those are different projects requiring different analyses, and the answer to "how many hours should it be" is only interesting in the first case. A developer that runs the duration study before establishing which constraint binds may have done a careful piece of work on a question its site does not ask.
Grid Code: The Constraint Written Into the Agreement
Less visible than the first three because it arrives as a schedule to the interconnection agreement rather than as a physical limit, and it can bind either variable.
The requirements that matter for a battery are mostly about behaviour rather than size: fault ride-through, reactive power capability across a defined range, frequency response characteristics, ramp rate limits, and the control and telemetry the operator requires.
Three of them have configuration consequences.
Reactive power capability consumes apparent power. An inverter rated at 100 MVA delivering reactive power is not simultaneously delivering 100 MW of real power. A requirement to hold reactive capability across the full operating range therefore reduces the real power available for the energy market, and the effect is larger at the edges of the range.
Ramp rate limits can cap the products you sell. A battery's commercial advantage is that it moves quickly. A ramp rate restriction narrows the gap between a battery and a conventional resource, and it directly limits participation in the fastest-responding products — which, as the Series C battery post established, are the ones storage is best suited to.
Minimum import or export thresholds create dead bands where the asset cannot operate economically, which matters more than it sounds for a resource whose value comes from responding to small price differences.
The practical instruction is to read the technical schedules to the interconnection agreement as commercial documents, because their effect lands entirely on the revenue side. A project that models 100 MW of energy-market participation against an agreement requiring reactive support has modelled a machine it has not agreed to build.
Offtake: The Constraint on What Is Worth Building
Distinct from the others because it does not limit what you can build. It limits what is worth building, which is the same thing commercially and a different thing physically.
A merchant battery can be any configuration its site allows. A contracted one is shaped by the contract.
A tolling agreement specifies the asset. The toller is buying dispatch rights over a defined power rating and duration, and the payment is calibrated to them. Building more than the toll contemplates produces capacity the toller does not pay for; building less is a breach. So in a tolled project, the offtake is the binding constraint on both variables at once — and it was set in a negotiation rather than by physics.
A capacity obligation sets a floor on duration. Where accreditation rules reward duration — the four, six, eight and ten hour ladder from the Series C accreditation post — a project intending to sell capacity has a duration decision made largely for it.
A hybrid or co-located arrangement imports the host's constraints. Sharing a point of interconnection with a solar array means the battery's available export capacity is whatever the array is not using, which is the subject of a later post in this series.
The order of operations matters here. Signing a toll before establishing the site's physical constraints risks committing to a configuration the site cannot deliver. Establishing the constraints first and then contracting to them is slower and avoids the worst outcome in development, which is a signed obligation against an unbuildable asset.
Supply Chain: The Constraint That Binds on the Calendar
The odd one out, because it rarely limits size and frequently limits timing.
Cell availability, container delivery slots and transformer lead times determine when a project can be built rather than how large it can be. For most of this series that would be a footnote. It is not, for one reason: the tax deadlines.
As the policy series established, a project's credit eligibility depends on dates — and under current rules wind and solar must begin construction on or before 4 July 2026 to preserve the four-year continuity safe harbour, or else be placed in service by the end of 2027. Storage sits under its own rules, but the structural point holds: a delivery slot that lands the wrong side of a statutory date is not a scheduling inconvenience, it is a credit eligibility event.
That makes procurement timing a financial constraint wearing an operational costume. The right treatment is to model the delivery schedule against the date ladder rather than against the construction programme, and to treat a slot that clears the tax date as worth paying a premium for — because the premium is bounded and the credit is not.
The Constraint You Can Buy, and the Ones You Cannot
Worth separating, because it changes the negotiation.
Buyable: land (more acreage, at a price), capital budget (more equity or debt, at a cost), and to a degree supply chain (pay for an earlier delivery slot).
Not buyable: interconnection capacity in any reasonable timeframe, fire code requirements, and grid code. These are the hard edges of the problem.
The practical consequence is that a project should be optimised against the non-buyable constraints, treating the buyable ones as prices rather than limits. A developer that treats its capital budget as a hard constraint and the interconnection as negotiable has inverted the problem — the budget is a decision and the interconnection is a fact.
How Do You Model the Constraint Stack in Excel?
As a table of limits with the binding one identified and the shadow price of each reported.
The constraint table
Limit on power Limit on energy
Interconnection 100 MW —
Land (20 acres usable) — 870 MWh
Fire code spacing — (in the land figure)
Capital budget $110m varies varies
Grid code 100 MW —
Binding on power: Interconnection at 100 MW
Binding on energy: Capital budget
The budget constraint, which is a line not a point
The capital budget does not limit power or energy on its own — it limits a combination, and the trade-off is set by the cost structure from the next post but one:
Budget = Power_MW × BOS_$/kW × 1,000 + Energy_MWh × Pack_$/kWh × 1,000
$110,000,000 = MW × 380,000 + MWh × 180,000
At 100 MW that leaves (110,000,000 − 38,000,000) / 180,000 = 400 MWh, or exactly four hours. At 80 MW it leaves 442 MWh — 5.5 hours on a smaller power rating.
That line is the actual decision space, and plotting it is the single most useful thing this analysis produces.
The shadow price of each constraint
Shadow price = value of relaxing the constraint by one unit
Interconnection: value of one more MW of POI
Land: value of one more acre
Budget: value of one more dollar of capex
A constraint with a shadow price of zero is not binding and can be ignored. A constraint with a large shadow price is where effort should go — and if that constraint is the interconnection, the honest conclusion may be that the effort belongs on a different site.
The screening test that comes before all of this
For a developer looking at a pipeline of candidate sites rather than a single project, the useful output is not a shadow price at all. It is a rejection rule.
For each candidate site:
Distance to usable substation → reject beyond threshold
Available interconnection capacity → reject below minimum viable MW
Usable acreage after setbacks → compute max MWh
Implied duration at max MW = max MWh / interconnection MW
If implied duration < 2 hours → energy-constrained, likely uneconomic
If implied duration > 8 hours → power-constrained, consider a larger connection
That last ratio is the single most useful screening number in storage development, and it takes two inputs a developer already has. A site whose interconnection and usable land imply a ten-hour battery is telling you the connection is too small for the parcel; one implying ninety minutes is telling you the parcel is too small for the connection. Either is a site worth reconsidering before any revenue modelling happens at all.
ℹ️ Note: Report which constraint binds as a model output, not an assumption. A model told that the project is 100 MW / 400 MWh cannot tell you whether that configuration was chosen or merely inherited, and the difference is usually worth more than anything else in the analysis.
To build the constraint table, the budget frontier and the shadow prices, prompt Dezzmond with your site, interconnection and budget parameters.
What Do Developers Actually Check?
- Which constraint binds on power, and which on energy? They are usually different.
- How far is the site from a usable substation? It eliminates more sites than anything else.
- Is the interconnection service full or conditional, and is curtailment acceptable for a battery?
- What is the usable acreage after setbacks, not the parcel size?
- Does the equipment vendor have large-scale fire test results supporting tighter spacing?
- Is the capital budget being treated as a constraint or as a price? It is a price.
- What is the shadow price of each constraint, and is effort going to the largest?
Frequently Asked Questions
What limits the size of a battery project?
Different things limit different dimensions. Interconnection capacity limits power; land and fire code limit energy; the capital budget limits a combination of the two. Finding which binds first determines which analysis is worth doing.
Why does interconnection matter more than land?
Because substation proximity eliminates more candidate sites than acreage, terrain or zoning, and because interconnection capacity cannot be bought in any reasonable timeframe while land usually can.
How much land does a battery need?
Roughly 1,000 square feet per MWh for the containers themselves — about 2.3 acres for a 100 MWh project — with total site area including setbacks, access and transformer yard typically running 5 to 20 acres.
What changed in the 2026 NFPA 855 edition?
Hazard Mitigation Analysis became mandatory for virtually all installations over 1 kWh, and explosion control per NFPA 69 or a tested performance-based alternative is now required. Unit separation remains three feet unless large-scale fire testing supports less.
Does a tolling agreement constrain the configuration?
Yes, on both variables at once. The toller buys dispatch rights over a defined power rating and duration, and the payment is calibrated to them — so the offtake becomes the binding constraint, set by negotiation rather than by the site.
Why does supply chain matter if it does not limit size?
Because it limits timing, and timing is a credit eligibility question. A delivery slot landing the wrong side of a statutory beginning-of-construction or placed-in-service date is not a scheduling problem, it is a tax event.
Does adding duration require a new interconnection study?
Generally not, because duration does not change the power presented to the grid. That is why hours are the cheap dimension to expand on a power-constrained site.
Closing: Find the Edge Before You Optimise Inside It
The instinct in a development team is to start with a configuration, because a configuration is something you can model. The constraint stack is unglamorous by comparison — a list of limits, mostly discovered by asking other people.
But the configuration is downstream of the limits, and the limits are not interchangeable. An hour of duration is nearly free on a power-constrained site and impossible on a land-constrained one. A megawatt is available on a large parcel with a small connection and unobtainable on a small parcel with a large one. The same analysis produces opposite answers depending on an input that most models do not contain at all.
The discipline is one line of output: which constraint binds, on which variable, and what would it cost to relax. With that, the rest of this series is a set of tractable questions. Without it, the duration study, the dispatch model and the revenue forecast are all being run on a configuration nobody established was available.
The next post takes the first of those variables on its own terms: what the megawatts are actually for, and how to size the power rating against the revenues that pay for it.
Sources: Telgian — NFPA 855 Changes in the 2026 Edition · Exponent — NFPA 855 Expands Safety Guidelines for Battery Energy Storage Systems · Sunnyplans — Land Requirements for BESS Projects: What's Different from Solar · Mayfield Renewables — NFPA 855 ESS Unit Spacing Limitations