EPC

What it is (precise)

An EPC contract bundles three scopes that are otherwise separate. Engineering covers the detailed plant design: DC block and enclosure layout, power conversion system (PCS) sizing in MVA, MV collection and transformer design, protection and SCADA, fire detection and suppression, civil and structural foundations, grounding, and the interconnection facilities up to the POI.

Procurement is the sourcing and supply-chain management of containerized battery enclosures, PCS, transformers, switchgear, cabling and balance-of-plant. Construction is the physical build, installation, integration, testing and commissioning. The contract is usually structured as a lump-sum turnkey (LSTK) agreement so that cost and schedule risk transfer largely to the contractor, and the scope boundary is normally the POI or the high side of the main power transformer.

Two structures carry the name. In a full EPC wrap, one contractor delivers the whole plant — batteries, PCS, balance of plant — under a single contract with unified guarantees: rather than the owner coordinating a battery OEM, a civil contractor and an electrical contractor under several contracts, the EPC is contractually liable for the integrated result, and everything from cell to grid inside the fence line is its problem until handover.

In a split, the owner buys the battery system directly from the OEM under a Battery Energy Supply Agreement (BESA), contracts balance-of-plant and installation separately — sometimes called EPC-lite or a BOP contract — and coordinates the interfaces itself. The label survives in both arrangements; what changes is what the contractor actually procures and what its guarantees stand behind.

The EPC wrap: what a turnkey contract buys, and what it costs

The wrap exists largely to make a project financeable. Lenders sizing debt against a Power Purchase Agreement or Tolling agreement want a defined cost, a defined completion date, and one accountable party who carries integration risk and backs it with performance guarantees, liquidated damages and warranties.

Because the borrower is usually a thinly capitalized Special Purpose Vehicle with no balance sheet of its own, the EPC wrap is one of the main things standing behind the debt: a creditworthy contractor with a full wrap directly improves Bankability and lowers the cost of capital at financial close.

The premium is the price of that risk transfer — the contractor stacks margin and contingency on every subcontract and equipment price it wraps, including, in a full wrap, the battery system itself — and the wrap is only as good as the contractor's balance sheet, because a guarantee from a firm that cannot absorb the loss is decoration.

What the wrap buys is singularity: one design authority, one schedule, one liquidated-damages regime, one party to pursue. Nothing falls between contracts because there is only one contract, and the owner's project team can be small. What it costs, beyond the premium, is visibility and optionality. The owner sees a lump sum rather than the equipment prices inside it, and cannot easily re-tender the DC blocks when their pricing moves — which on this asset class it does, quickly.

Cell and integrator selection becomes the contractor's commercial decision constrained by the owner's specification, so the specification has to be right the first time and enforceable in the technical schedules. And the contingency priced into the wrap is spent whether or not the risk it covers ever occurs.

The split: a BESA plus a BOP contract, and where it bites

The BESS-typical split is a battery supply agreement signed directly with the OEM plus a balance-of-plant EPC — the common structure on this asset class. The owner buys what is usually the largest single cost item at OEM pricing with no intermediary margin, and gets warranty privity: capacity, efficiency and defect claims run straight to the manufacturer instead of through a contractor that may not exist when the defect surfaces.

It also keeps the integrator honest, because the owner now holds two commercial relationships and can see what each scope actually costs. Timing helps too: a battery package can be re-tendered or re-specified later in development than a wrapped price allows.

That premium arithmetic is why the split became the mainstream structure for large BESS projects: DC block prices fell fast enough that owners wanted direct OEM relationships and OEM pricing, and paying an integrator's margin and contingency on top of what is usually the plant's dominant cost item stopped making sense. The saving is real; the cost is that integration risk moves to the owner.

If the plant misses its capacity test, the battery OEM points at the PCS supplier, who points at the EPC, and the owner must prove whether equipment or installation is at fault before either warranty responds — which is why the interface between the OEM's equipment warranty and the EPC's workmanship warranty, especially around capacity retention, round-trip efficiency, availability and code compliance for fire and explosion protection, becomes the most important seam the owner manages. Lenders price that gap, so the cheaper structure is not automatically the better one for Bankability.

Running a split takes a capability, not just a decision. The owner needs an owner's engineer who can arbitrate technical disputes between two vendors on the day they happen, a division-of-responsibility matrix maintained as a live document, and schedules interlocked at delivery dates rather than at milestones each party sets for itself.

Check the damages regimes add up: a BESA's late-delivery damages are set against the equipment price and a BOP contract's delay damages against the construction price, and neither is calibrated to the owner's revenue loss under the offtake. The difference between the sum of those two remedies and the owner's actual exposure is retained risk, whether or not anyone books it.

Exclusions, the DOR matrix, and the tests that close the contract

Read the exclusions before the scope. A BESS EPC scope commonly excludes everything beyond the POI (utility-side works and network upgrades, which the interconnection agreement allocates on its own terms), permits and land rights, ISO registration and market qualification, unknown ground conditions handled through a differing-site-conditions clause, offloading and storage of owner-furnished equipment, long-term spares, and the fibre or telemetry path back to the ISO.

Every excluded item still has to happen. The discipline is to walk the exclusions list and write a name against each one — the owner, the OEM, the utility, a separate contractor — because an exclusion that nobody claims becomes a change order priced under duress, usually in the last month before energization.

A working engineer meets the EPC first as a division-of-responsibility (DOR) matrix and a scope book. Read the split carefully: who furnishes the battery enclosures and PCS, who performs the UL 9540A-based hazard mitigation analysis for the AHJ, who builds and validates the plant model for interconnection studies, and who is responsible for grid-code compliance tests at the POI under IEEE 2800 or the utility's requirements.

Owner's engineers spend much of their review time on exactly these seams, because any function that appears in neither column — auxiliary power sizing, thermal management ducting, SCADA points lists — becomes a change order later.

Milestones then structure everything. Mechanical completion leads to commissioning per NFPA 855, then a capacity test and a round-trip-efficiency test at substantial completion.

Whether that certificate also satisfies the conditions for the Commercial Operation Date under the Power Purchase Agreement or Tolling agreement is a drafting question, not a rule: some stacks define substantial completion to require the capacity test, permission to operate and market registration, so the two land together; in others those conditions stay open for weeks or months after the contractor has finished and delay damages have stopped.

Check the test protocol details: is guaranteed capacity stated in MWh AC at the POI or DC at the battery terminals, at what ambient temperature and power factor, with auxiliary load netted or excluded, is RTE defined AC-to-AC including auxiliary loads or DC-to-DC excluding them, and whose revenue meter governs?

A guarantee at the inverter terminals at 25 C and unity power factor is far weaker than the same number at the POI on a hot day, and a 400 MWh DC nameplate can be materially less at the POI after PCS, transformer and auxiliary losses. Availability guarantees — high-90s percent is typical territory — turn entirely on how "available" is defined and measured, so those definitions deserve senior attention before signing, not after the test fails.

Typical values and standards

Whoever supplies the batteries, the EPC designs and builds to the same code stack. The US installation standard is NFPA 855, which sets separation distances, maximum stored energy per group, and commissioning requirements. NFPA 855 references UL 9540 (the system-level safety listing for energy storage equipment) and the UL 9540A test method, which characterizes thermal-runaway fire-propagation behavior — AHJs typically require UL 9540A cell-, module- and unit-level test reports to set spacing and fire-protection measures before permitting.

Where deflagration is credible, NFPA 68 (deflagration venting) and NFPA 69 (explosion prevention) apply to enclosure protection. IEEE 1547 (distribution) and IEEE 2800 (transmission), plus the utility's interconnection agreement, govern grid behavior at the POI. Who furnishes the test reports behind that stack is a scope question, and it must land in someone's column.

On cost, treat every figure as a range. Mid-2020s US utility-scale systems of 2–4 hours have commonly landed somewhere around $200–400 per kWh installed all-in, shorter durations sitting toward the top of the band because the power block and balance of plant spread over less energy. Inside that, the battery system is usually the single largest line item at very roughly 35–45% of capital cost, the power conversion system another 15–20%, and the EPC and balance-of-plant scope typically 30–40% when the batteries are owner-furnished.

Where the connection requires its own high-voltage scope, that adds a further 20–25% on top, not a share of the three. These splits feed straight into any levelized-cost model, so an estimator should always ask which scope a quoted $/kWh actually covers — a figure that silently excludes owner-furnished batteries describes a different plant than one that wraps them.

Notice to proceed to COD commonly runs on the order of 6–18 months depending on plant size, site conditions and interconnection readiness — and the last 2–4 months of that are commissioning, capacity testing and market qualification rather than construction.

Contract levers carry typical numbers too, and every one of them is a negotiated position rather than a standard — FIDIC forms, US-style EPC forms and owner-drafted forms allocate these differently. Delay liquidated damages often run on the order of 0.05–0.15% of contract price per day, with an aggregate LD cap commonly in the 10–20% range — on a $200 million contract at 0.1% per day capped at 10%, that is $200,000 per day exhausting after 100 days, with revenue at risk above the cap self-insured whether the owner realizes it or not.

Performance LDs buy down shortfalls against guaranteed capacity, round-trip efficiency and availability. Retainage of roughly 5–10% is withheld until completion milestones, backed by performance bonds or letters of credit often sized near 10% of contract value. A defects-liability period of 1–2 years on workmanship is standard, sitting alongside much longer OEM warranties — typically 10–20 year capacity and degradation guarantees on the battery blocks that the EPC explicitly does not wrap.

Common pitfalls

The first trap is reading the label as the substance. An "EPC" whose contract excludes the owner-furnished batteries wraps far less than the name implies: its guarantees cover balance-of-plant and workmanship, while the capacity and degradation promises that dominate the project's risk sit in the BESA with the OEM. Before relying on a wrap, establish what the guarantees actually cover, at which measurement boundary, and against whose balance sheet — then check that no obligation the owner owes its offtaker or lender falls between the two contracts.

Two quieter trip-wires. First, calendar aging: cells degrade between factory test and energization, so a capacity guarantee referenced to nameplate can fail at the COD test if the contract ignores months of transit and construction storage — guarantees should state beginning-of-life values at the test date, with degradation explicitly allocated.

Second, code compliance is not paperwork: an EPC that treats NFPA 855 spacing or deflagration protection as an afterthought creates permitting delay, retrofit cost and impaired Insurability, since property insurers increasingly underwrite against UL 9540A data and site layout. Asking who signs the hazard mitigation analysis is a fast way to test an EPC's BESS maturity.