Sizing Utility-Scale BESS: Choosing Between 1-Hour, 2-Hour and 4-Hour Duration
Duration selection methodology for utility battery storage: revenue fit, power ratio economics and sizing workflow for 1, 2 and 4 hour assets.
Duration selection methodology for utility battery storage: revenue fit, power ratio economics and sizing workflow for 1, 2 and 4 hour assets.
A dispatch-modeling approach to storage duration selection: what one-hour, two-hour and four-hour assets each earn, what they cost, and how to find the ratio your revenue contract actually rewards.
Duration is the defining design choice on a storage project, more consequential than chemistry and less reversible than vendor selection. A one-hour asset is a power device: it sells speed, frequency response, and ramping capability, with energy capacity sized merely to support bursts. A four-hour asset is an energy device: it sells shifted megawatt-hours across the evening peak, with power capability sized to move them. The two-hour asset sits at the inflection, capable of both while optimizing neither, and Indian procurement—spanning 2-hour rounds like the SECI 500 MW / 1,000 MWh ESS-I tender and 4-hour tranches like the SECI 800 MW / 3,200 MWh ESS-IV tender—now forces the duration choice explicitly.
The grid problem behind the choice is concrete. Under the Central Electricity Authority (CEA) National Electricity Plan (2022–2032), India projects 47.24 GW / 236.22 GWh of BESS alongside 26.69 GW / 175.18 GWh of pumped storage by FY 2031–32 to integrate 364 GW of solar and 121 GW of wind. That renewable profile creates an evening peak coinciding with declining solar output—a ramp requirement measured in tens of gigawatts over a few hours—alongside frequency response obligations governed by the CERC (Ancillary Services) Regulations, 2022. One problem rewards duration; the other rewards power. A single asset can serve both only by carrying cost structure sized to the harder of the two, and that cost structure is what duration selection arbitrates.
The economics follow the power-to-energy ratio through the cost stack, as structured in the NREL Annual Technology Baseline (ATB) and NREL Cost Projections for Utility-Scale Battery Storage. Conversion equipment scales with power (USD per kW); cells scale with energy (USD per kWh); balance of plant splits between them. A one-hour asset carries roughly four times the conversion hardware per megawatt-hour of a four-hour asset, which is why its cost per megawatt-hour looks punishing while its cost per megawatt looks reasonable. Neither number means anything alone: the revenue model decides which denominator is real.
This article sets out the sizing methodology our practice applies: characterize the duty from the revenue contract, model dispatch against price and grid signals, extract the duration histogram the duty actually demands, and size power and energy against it with margin analysis. It compares the three duration archetypes on revenue fit and cost structure, and closes with the workflow that converts the analysis into a defensible tender specification.
The one-hour asset earns from speed markets. Primary, Secondary (AGC-dispatched SRAS), and Tertiary Reserve Ancillary Services under the CERC Ancillary Services framework, along with dynamic response and operating reserves, pay for capability delivered in seconds to minutes, and the energy content of each delivery is small. While central and state standalone utility tenders in India have predominantly tendered 2-hour and 4-hour capacities, 1-hour configurations suit high-power frequency regulation and industrial fast-response applications. The risk profile is regulation-market reform: rules, settlement intervals, and qualification requirements change, and a one-hour asset has nowhere to hide from them because it has no second revenue line sized to carry it.
The four-hour asset earns from shape. Evening peak shaving, renewable firming across the sunset ramp, and capacity obligation products pay for sustained delivery across hours, and the asset competes against gas peakers and demand response on delivered shape rather than speed. The revenue is contracted and stable in most Indian structures, which is why four-hour assets anchor the financed end of the market. The risk profile is forecast: load shape evolution and renewable buildout move the evening spread the asset depends on.
The two-hour asset earns from optionality. It serves ramping duty, the two-to-three hour sunset window in India, and capacity-style tenders written at two-hour duration, while retaining partial regulation capability. Indian procurement has written substantial two-hour capacity into its programs precisely because the national ramp sits in that band. The risk profile is mediocrity: when regulation prices spike, the two-hour asset lacks the power headroom to capture them fully, and when spreads extend across four hours, it empties before the revenue window closes.
Duration and cycling intensity interact with fade, which our degradation analysis covers. Four-hour assets cycle deeply but once daily; one-hour assets cycle shallowly but continuously, aging through resistance growth and thermal ripple instead. The augmentation arithmetic differs by archetype, and the sizing decision should carry fade modeling matched to the cycling signature, not a generic degradation percentage applied uniformly across the candidates.
Sizing begins with the revenue contract, not the market report. Parse the contracted obligations, availability windows, response requirements, and energy delivery commitments into a duty specification, then layer merchant opportunities on top as sensitivities rather than base case. The most common sizing error inverts this: models built on optimistic merchant stacking produce specifications that cannot service the debt when merchant revenue underperforms, and the asset spends its life too small for its contract and too large for its market.
Dispatch modeling converts the duty into numbers. Against the price curves, grid signals, and contract windows, the model dispatches candidate assets across representative historical periods, and the output that matters is the duration histogram: how much energy the asset would deliver per event, at what power, across how many events, at what state-of-charge depths. The histogram, not the peak demand figure, is what the asset must be sized to.
Power and energy are then extracted from the histogram with margin analysis. Power sizing follows the highest-revenue sustained delivery window, energy sizing follows the depth distribution that captures the contracted energy at acceptable state-of-charge ceilings, and both carry margin against forecast error sized to the revenue at risk. The margins are economic choices, not engineering conservatism: margin costs capital, and the correct amount is the amount that protects the revenue distribution the financing depends on.
The final step reconciles against tender structure. Where the tender prescribes duration bands or minimum energy commitments, the optimization runs within them, and partial compliance options, where permitted, deserve explicit modeling. Where the tender leaves duration open, the histogram-driven ratio is the bid specification. In both cases the deliverable is the same: a sizing memorandum tying power, energy, and margin decisions to the dispatch evidence, auditable by the lender and defensible in diligence.
The comparison below scores the three archetypes on revenue fit, cost structure, and risk profile as they present in the Indian market. Cost shares are structural: conversion hardware scales with power, cells with energy, and the balance splits. The table is a decision instrument for the histogram output, not a substitute for it, and the final column names the tender context that typically selects each archetype.
Reading the table against the cost analysis in our cost guide: the per-megawatt-hour capital ranking inverts the per-megawatt ranking, which is the entire point. Bids evaluated on cost per megawatt systematically favor long-duration assets; bids evaluated per megawatt-hour systematically favor short ones; and only the levelized comparison against the revenue histogram is honest between them.
| Dimension | 1-Hour Asset | 2-Hour Asset | 4-Hour Asset | Deciding Tender Context |
|---|---|---|---|---|
| Primary revenue line | Frequency and response markets | Ramping and capacity tenders | Peak shifting and firming | Contract structure |
| Conversion cost per MWh | Highest | Moderate | Lowest | Power-to-energy ratio |
| Cell cost per MW | Lowest | Moderate | Highest | Duration of delivery |
| Cycling signature | Shallow, continuous | Moderate depth, daily | Deep, once daily | Fade modeling basis |
| Contracted revenue stability | Market-rule dependent | Tender-backed | Tender-backed | Counterparty structure |
| Merchant upside capture | Speed markets only | Partial both sides | Spread markets only | Stacking ambition |
| Ancillary qualification | Full capability | Partial capability | Duration-limited capability | Grid code provisions |
| Augmentation profile | Resistance-driven fade | Mixed mechanism | Calendar and depth fade | Warranty structure |
| Typical Indian tender fit | Regulation pilots, DSMs | State capacity rounds | National capacity programs | Program alignment |
Real assets face stacked duty, and the ratio decision becomes a frontier optimization: mapping revenue combinations against ratio candidates and finding where the marginal megawatt of power and the marginal megawatt-hour of energy each stop paying. The frontier has a shape: too little power and speed revenue is capped; too little energy and duration revenue is truncated; and the optimum sits where the two constraints bind simultaneously, which the dispatch model locates numerically.
Stacking discipline protects the financing. Contracted revenue funds the debt, so the ratio must first satisfy the contracted duty with margin, and merchant overlays, regulation participation, trading windows, and any merchant energy sales, are then modeled as incremental revenue on the residual capability. An asset sized to serve merchant upside as if it were contracted is financing itself on revenue that is not, and Indian lenders diligence exactly this boundary when they diligence anything.
The frontier moves with market structure, and the sizing decision should carry that movement as sensitivity rather than certainty. Regulation market reform shifts value toward power; renewable buildout and load growth extend evening spreads toward duration; transmission expansion changes congestion patterns that local assets monetize. The robust choice is the ratio whose economics survive the plausible range of these shifts, and the sizing memorandum should show that survival explicitly, because robustness, not optimality, is what twelve-to-fifteen-year Indian storage financing actually requires.
The nameplate error sizes duration to the peak demand hour rather than to the delivery histogram, producing assets that empty ninety minutes into four-hour revenue windows. The cost is silent: availability statistics look healthy while the revenue that depends on the final hours of the window never materializes, and the shortfall surfaces in the reforecast that lenders request when covenant results drift.
The stacking error sizes to merchant optimism. Its signature is a bid memo whose revenue lines assume regulation participation, trading windows, and contracted capacity simultaneously at full intensity, with no curtailment of one revenue line by another. Physical reality arbitrates: an asset committed to capacity availability cannot trade that energy, and the model that never let the lines compete with each other prices an asset the duty cannot deliver.
The symmetry error assumes charging and discharging power must match. Asymmetric ratios, higher discharge than charge power or the reverse, track the duty histogram when generation and load sides of the market differ, and where tender rules permit them, they buy real economics. The error costs little by itself but signals the underlying failure: a sizing process that never interrogated the histogram at all.
The margin error runs both directions, and both are expensive. No margin leaves the asset hostage to forecast error in its contracted duty, converting a two-percent load-shape miss into a covenant conversation. Uniform margin, applied as a blanket percentage rather than sized to revenue at risk, wastes capital on windows that carry no risk while under-protecting the ones that carry all of it. Margin is a priced instrument in mature sizing practice, allocated where the revenue distribution says it earns.
The workflow below condenses the methodology into a gated sequence with named deliverables. Its discipline is evidence-chain: every sizing decision traces to a dispatch artifact, so the lender diligence that follows financial close reviews evidence rather than re-litigates assumptions. The workflow is tender-agnostic, applying equally to open-duration bids and to banded tenders where the optimization runs within the band.
The final deliverable is the sizing memorandum, and its structure should be standardized across a portfolio: duty specification, dispatch evidence, duration histogram, ratio derivation with margin analysis, fade and augmentation implications, and sensitivity band. A developer holding standard memoranda across projects builds a reusable evidence library that accelerates every subsequent bid, which at Indian tender cadence is a durable competitive position.
Portfolio consistency adds the last increment of value. Sites sized under one methodology can be re-banded, re-augmented, and re-financed against one another as market structure shifts, because their evidence chains share structure and vocabulary. Sites sized ad hoc cannot, and the difference surfaces at exactly the moments portfolios need flexibility most: tranche consolidation, refinancing windows, and portfolio-level trading participation.
| Workflow Gate | Core Activity | Deliverable | Common Shortcut | Detection at Diligence |
|---|---|---|---|---|
| 1. Duty specification | Parse contract into obligation envelope | Duty spec document | Market report as proxy | Revenue lines mis-summed |
| 2. Dispatch modeling | Simulate candidates on representative periods | Dispatch logs and revenue stack | Single representative day | No seasonal evidence |
| 3. Duration histogram | Extract delivery depth and power distribution | Histogram with percentile bands | Peak-hour assumption | Final-hour revenue gap |
| 4. Ratio derivation | Fix power and energy with margin analysis | Ratio memo with sensitivity | Symmetric default ratio | Histogram mismatch |
| 5. Fade alignment | Match degradation model to cycling signature | Fade annex to sizing memo | Uniform fade percent | Augmentation unpriced |
| 6. Tender reconciliation | Verify compliance with duration rules | Compliance checklist | Band minimum assumed | Disqualification or repricing |
Central and state utility standalone tenders have split predominantly between two-hour configurations (such as SECI's 500 MW / 1,000 MWh Fatehgarh-III and GUVNL Phase IX 450 MW / 900 MWh rounds targeting the sunset ramp) and four-hour configurations (such as SECI's 800 MW / 3,200 MWh ESS-IV and GUVNL Phase X 1,000 MW / 4,000 MWh rounds anchored on evening peak shifting), while CEA's National Electricity Plan 2031–32 projection (47.24 GW / 236.22 GWh) implies an average duration approaching five hours. Sub-hourly and one-hour capability maps primarily to ancillary services under CERC's SRAS/TRAS framework rather than standalone bulk-capacity tenders.
Only through the tender compliance rules rather than through dispatch physics. Some structures permit aggregated or partial compliance, and a portfolio of two-hour assets can serve a four-hour obligation through coordination if the rules, metering, scheduling, and settlement, allow it. The dispatch physics do not negotiate: a two-hour reservoir cannot deliver a four-hour window without recharge inside it, and any bid model assuming otherwise will fail its own availability arithmetic at the first test.
Through the cycling signature. Four-hour assets cycle deeply once daily, aging through depth and calendar mechanisms with near-linear fade. One-hour assets cycle shallowly but continuously, aging through resistance growth and thermal ripple, which surfaces first as power capability loss rather than capacity loss. Two-hour assets mix both signatures. Augmentation tranches should therefore be modeled against the mechanism mix of the chosen archetype, or the schedule prices the wrong intervention at the wrong time.
Where tender rules permit it, yes, when the duty histogram is genuinely asymmetric: generation-side charging patterns and load-side delivery windows frequently are. Discharge power sized above charge power captures evening delivery peaks the symmetric ratio would truncate, and the reverse suits assets charging from constrained renewable surplus. The specification costs little at design time and buys real revenue. The caution is operational: asymmetric hardware complicates compliance demonstration under CEA grid connectivity standards, and the witness testing plan should account for it explicitly.
Take the contracted energy obligation per delivery window, divide by the window length, and compare that implied power against the revenue-weighted histogram peak. If the two numbers differ by more than about twenty percent, the duty and the obligation are in tension, and full modeling is needed before any specification is written. If they align, the ratio is probably near the histogram optimum already, and modeling effort can focus on margin and sensitivity rather than on discovering the shape.
Ultrathon Electric's multidisciplinary engineering division prepares technical analyses, grid-code evaluations, and safety briefs to assist utilities, independent power producers (IPPs), and investors in navigating utility-scale energy storage and renewable deployments.
Whether you are an IPP bidding for SECI / State DISCOM storage tenders or a large-scale commercial consumer requiring peak shaving and power backup, Ultrathon Electric provides end-to-end BESS EPC execution and Owner's Engineer advisory.
Northern Solar Assets secured 189 MWac CRESS solar-plus-BESS LOAs for two Malaysian data centres. Verified disclosures, gaps and engineering diligence.
Market IntelligenceUnion Cabinet approves Rs 14,500 crore VGF scheme for 92 GWh grid-scale storage (30 GWh BESS, 60 GWh PSP, 2 GWh emerging) with 40% BESS DCR.
Need customized techno-commercial modeling, single-line diagram reviews, or tender specification advisory?