When Should a Battery Charge? The Threshold Is Degradation, Not Efficiency
A battery is about to charge at $20/MWh. What does that cycle cost?
Round-trip efficiency loss (88%) $2.73/MWh
Degradation $30.00/MWh
Degradation is eleven times the efficiency loss, and it is the cost that almost every model treats as an annual capacity haircut rather than as a price on the decision in front of it.
That inversion matters because it changes the operating rule. Efficiency is usually presented as the cost of cycling — it is the number in every specification sheet and the one dispatch models are careful about. It is also, at realistic prices, a rounding error against the cost of consuming warranty throughput.
This post builds the charge decision properly: what a cycle actually costs, what minimum spread justifies one, why that threshold is roughly constant in dollars rather than in percentage, and where the honest uncertainty sits.
ℹ️ Note: Figures are labelled assumptions. The degradation cost in particular depends on warranty terms and on how much of the pack's life is cycle-driven rather than calendar-driven — a distinction this post takes seriously below.
What Is the Charge Decision?
Whether to buy energy now in order to sell it later. It is not a view on price levels; it is a view on the spread between now and the best available later, net of what the round trip costs.
A battery that charges at $20 and discharges at $25 has lost money. A battery that charges at $200 and discharges at $260 has made money. The price level is irrelevant; only the gap matters, and the gap has to clear a threshold.
The Naive Rule, and Why It Fails
The rule most people state first is a price threshold: charge below X, discharge above Y.
It fails for three reasons.
It is not robust to price levels. A rule calibrated on a $30 average market breaks entirely in a $90 one, and markets move.
It ignores the opportunity cost of the energy already stored — the subject of the next post.
It gets the cost of cycling wrong, which is what this post is about.
The correct form is not a price at all. It is a condition on the spread:
Charge at price P_c only if there exists a later hour with price P_s such that:
P_s ≥ P_c / RTE + Degradation_Cost_per_MWh
Read that as: what you sell must cover what you bought, grossed up for the energy lost on the round trip, plus the cost of the throughput you consumed.
The Two Costs of a Cycle
Round-trip efficiency is the energy that does not come back. At 88%, delivering 1 MWh requires importing 1.136 MWh, so the cost is the charge price times (1/RTE − 1):
At $20/MWh charge: 20 × (1/0.88 − 1) = $2.73/MWh
At $40/MWh charge: 40 × (1/0.88 − 1) = $5.45/MWh
Note that this cost scales with the charge price. Cheap energy is cheap to lose.
Degradation is the consumed fraction of the asset's throughput warranty. If the pack costs $180/kWh and the warranty permits 6,000 full cycles, each megawatt-hour discharged consumes:
$180,000 per MWh of capacity ÷ 6,000 cycles = $30.00/MWh
That cost is constant. It does not care what the energy cost or what it sold for; it is the price of using up a finite resource.
| Warranty throughput | Degradation cost per MWh discharged |
|---|---|
| 4,000 cycles | $45.00 |
| 6,000 cycles | $30.00 |
| 8,000 cycles | $22.50 |
The warranty is therefore a direct input to the dispatch rule, not merely a maintenance constraint. A tighter throughput allowance makes every cycle more expensive and raises the threshold at which cycling is worth doing.
The Minimum Viable Spread
Put the two together and the threshold emerges:
| Charge price | Required sell price | Required spread | As % of charge price |
|---|---|---|---|
| $10 | $41.36 | $31.36 | 314% |
| $20 | $52.73 | $32.73 | 164% |
| $30 | $64.09 | $34.09 | 114% |
| $40 | $75.45 | $35.45 | 89% |
The fourth column moves wildly. The third barely moves at all.
The minimum viable spread is roughly constant in dollars — about $31 to $35 per megawatt-hour across this range. That is because the degradation term is fixed and dominates, while the efficiency term, which does scale, is small.
This is a genuinely useful operating result. It means the rule can be stated as a single dollar figure rather than as a percentage or a pair of price thresholds, and that figure is stable as price levels move.
Rule of thumb on these assumptions:
Do not cycle for less than about $33/MWh of spread.
Everything below that destroys value, however attractive the percentage looks. A charge at $10 and a discharge at $30 is a 200% gain on the energy and a loss once the throughput is priced.
The Honest Uncertainty: Calendar Versus Cycle
The $30 figure is an upper bound, and the reason is worth understanding because it moves the threshold materially.
A battery degrades two ways. Cycle degradation is caused by use. Calendar degradation happens with time regardless of whether the asset cycles at all — a battery sitting idle for a year still loses capacity.
Only the first is a marginal cost of the decision to cycle. Calendar degradation is going to happen whether or not you charge tonight, which makes it a sunk cost in the dispatch decision, exactly as the fixed operating costs are.
So the marginal degradation cost is:
Marginal_Degradation = Pack_Cost ÷ Warranty_Cycles × Cycle_Driven_Share
At a 60/40 cycle/calendar split: $30.00 × 0.60 = $18.00/MWh
At an 80/20 split: $30.00 × 0.80 = $24.00/MWh
And the threshold moves with it:
Minimum spread at $30/MWh degradation ≈ $33/MWh
Minimum spread at $18/MWh degradation ≈ $21/MWh
A 35% difference in the operating threshold, from an assumption most models do not contain at all. A battery operated on the conservative figure is leaving cycles on the table; one operated on an optimistic figure is consuming warranty faster than it is being paid for.
Two practical points.
The split is knowable. Warranty documents typically specify capacity retention against both time and throughput, and the two curves together imply the split. It is a document already in the data room.
Residual value cuts the same way. A pack with meaningful residual or second-life value at end of warranty has a lower effective consumption cost than one written to zero, which lowers the threshold further.
Charging at Negative Prices Inverts the Arithmetic
The case where the efficiency loss stops being a cost and becomes a benefit — and it is not a curiosity, given that Series C documented over 1,200 hours of negative prices at ERCOT hubs in a single year.
When the charge price is negative, the battery is paid to consume energy. And because round-trip inefficiency means it must consume more energy to deliver a given amount, it gets paid more:
Charge at −$20/MWh:
Energy imported per MWh delivered 1.136 MWh
Payment received 1.136 × $20 = $22.73
Effective cost of the stored MWh = −$22.73
The threshold then becomes remarkably permissive:
| Charge price | Effective cost | Required sell price |
|---|---|---|
| −$20 | −$22.73 | $7.27 |
| −$10 | −$11.36 | $18.64 |
| $0 | $0.00 | $30.00 |
| $10 | $11.36 | $41.36 |
At a −$20 charge price the battery only needs to sell above $7.27 to be ahead — because the negative price has already paid for most of the degradation.
Three consequences.
Poor round-trip efficiency is an advantage in negative hours. A less efficient battery imports more energy and is paid more for taking it. That is the only circumstance in which the specification sheet's headline number works in the wrong direction, and it is worth knowing exists.
Negative-price hours are the highest-value charging opportunities, by a wide margin, and they cluster exactly where Series C said they would — in congested pockets with heavy renewable penetration. That is the same locational analysis, arriving at the storage siting decision from the opposite direction.
The degradation floor still binds. Even at very negative prices the asset must sell above roughly $7 to cover throughput. Charging at −$50 and discharging at $2 still loses money, which is counterintuitive enough to be worth stating in an operating procedure.
The Cycle Budget: When the Threshold Rule Is Not Enough
A second constraint that can override everything above, and it turns dispatch from a threshold test into a rationing problem.
A 6,000-cycle warranty over a fifteen-year life permits an average of 400 cycles a year. If the price series offers more than 400 opportunities that clear the minimum spread, the threshold rule will consume the warranty early.
Qualifying opportunities in the year say 520
Cycle budget 400
→ 120 profitable cycles must be declined
At that point the correct rule is not "take everything above the threshold." It is "take the best 400," which requires a different calculation: rank the opportunities and set the operating threshold at the 400th best spread rather than at the economic minimum.
Economic minimum spread ~$33/MWh
Budget-constrained threshold = 400th best spread in the year
Whichever is higher governs. And the budget-constrained threshold is only knowable from the full annual price distribution, which is another reason the representative-day approach fails for operating decisions.
Two further points.
The budget is not evenly distributed across the year. Spending it early because the first quarter offered good spreads leaves nothing for a volatile fourth quarter. Cycle budgeting is an intertemporal allocation problem, which is the same class of problem as the discharge decision in the next post.
Exceeding the budget is not free. Running more cycles than the warranty permits does not stop the battery; it voids or reduces the cover, which converts a warranty claim into an owner cost. The correct treatment is therefore not a hard constraint but a very expensive marginal cost above the limit.
The Threshold Is a Spread, Not a Price
Worth restating because it changes how the rule is implemented.
A price-based rule — "charge below $20" — embeds an assumption about what prices will be available later. When the market moves, the rule silently becomes wrong in one of two directions: too permissive in a high-price market, too restrictive in a low-price one.
A spread-based rule — "charge only if a later hour exceeds this price by $33" — is self-correcting. It needs no recalibration when price levels shift, and it directly encodes the economics.
The implementation requirement is that the rule needs a view of later prices, which is where forecasting enters and where the gap between a backtest and reality opens up. That is the subject of the final post in this series. For now the point is narrower: the decision variable is a spread against a known cost, and the cost is dominated by throughput rather than by efficiency.
The Competing Use: An Ancillary Award
One more term belongs in the threshold, and leaving it out is the most common way a dispatch rule overstates arbitrage.
A battery holding a frequency response or reserve award has committed its capacity to standing ready. It cannot simultaneously use that capacity to charge or discharge on price. So the true condition for cycling is not merely that the spread clears the cycle cost — it is that the spread clears the cycle cost plus whatever the ancillary award would have paid.
Charge only if: P_s ≥ P_c / RTE + Degradation + Foregone_AS_Revenue
That third term is an opportunity cost, and it is the reason the Series C battery post insisted that revenue streams cannot be added together. Energy and ancillary compete for the same megawatts and the same megawatt-hours; a stacked bar chart showing both at full value describes an asset doing two things at once.
Two practical observations.
The opportunity cost is the AS clearing price, not the AS revenue. What matters is what the capacity would have earned in that hour if committed to the other product, which is the marginal price rather than an annual average.
It has collapsed, which changes the rule. When ancillary services were paying well, the opportunity cost term dominated and batteries rationally sat on awards rather than arbitraging. With ERCOT ancillary revenues down roughly ninety percent since 2023, the term has shrunk and the same asset should now be cycling far more often than its original operating policy contemplated.
That is a specific, actionable point for anyone operating a battery configured in the ancillary era: the dispatch rule that was correct at commissioning is now too conservative, because one of its terms has fallen by an order of magnitude. An operating policy is not a set-and-forget document; it embeds market prices that move.
The general form is worth stating plainly. A battery's charge decision is a comparison between three uses of the same capacity — cycle now, cycle later, or stand ready for someone else — and the rule needs all three on the same scale.
How Do You Build the Charge Rule in Excel?
As a cost per megawatt-hour and a spread test, with both cost components separated.
The cycle cost
Efficiency_Cost(P_c) = P_c × (1/RTE − 1)
Degradation_Cost = Pack_$kWh × 1,000 ÷ Warranty_Cycles × Cycle_Share
Total_Cycle_Cost(P_c) = Efficiency_Cost(P_c) + Degradation_Cost
The threshold
Required_Sell(P_c) = P_c / RTE + Degradation_Cost
Min_Spread(P_c) = Required_Sell(P_c) − P_c
PF_MinViableSpread ≈ Degradation_Cost + P_c × (1/RTE − 1)
The decision
Charge_Now = AND(
P_c = MIN(prices in charge window),
MAX(prices in discharge window) ≥ Required_Sell(P_c),
SoC + charge ≤ SoC_max,
Cycles_used_this_period < Cycle_limit
)
The last two conditions are the subject of the next two posts, and they turn this clean rule into a scheduling problem.
The outputs to publish
PF_DegradationCostPerMWh $30.00 (or $18.00 marginal)
PF_EfficiencyCostAtTypicalPrice $2.73
PF_MinViableSpread ~$33.00 (or ~$21.00)
PF_HoursAboveThreshold count in the price series
That last line is the one that tells you whether the asset has a business at this node. A price series containing very few hour-pairs separated by more than the minimum spread describes a market where the battery should mostly sit idle — and no amount of dispatch optimisation changes it.
ℹ️ Note: Report the degradation cost assumption and its cycle/calendar split explicitly. It moves the operating threshold by more than a third, it determines how many cycles a year the asset should run, and it is the input most likely to be missing entirely.
To build the cycle cost, the minimum viable spread and the count of qualifying hours in your price series, prompt Dezzmond with your warranty terms, pack cost and node prices.
What Do Operators and Lenders Actually Check?
- What is the degradation cost per MWh, and what warranty throughput is it derived from?
- What share of degradation is cycle-driven rather than calendar?
- What is the minimum viable spread in dollars, and is the operating rule stated that way?
- How many hour-pairs in the actual price series clear it?
- Is the rule price-based or spread-based? A price rule breaks when levels move.
- Does the model price degradation into dispatch, or apply it only as an annual capacity haircut?
- Is there residual or second-life value reducing the effective consumption cost?
Frequently Asked Questions
What does it cost a battery to cycle?
Two things: the energy lost to round-trip inefficiency, which scales with the charge price, and the warranty throughput consumed, which does not. On the worked assumptions the second is around $30/MWh against $2.73 for the first at a $20 charge price.
What is the minimum spread worth cycling for?
Roughly $31 to $35 per megawatt-hour on these assumptions, and close to constant in dollar terms across a wide range of price levels — because the dominant component, degradation, is fixed.
Why is the threshold constant in dollars rather than percent?
Because degradation cost does not vary with the price of the energy. Only the efficiency component scales, and it is small — so the sum barely moves as price levels change.
Should calendar degradation be in the dispatch decision?
No. Calendar degradation happens whether or not the asset cycles, so it is sunk for the dispatch decision. Using total rather than cycle-driven degradation overstates the threshold by around a third.
What happens when charging at a negative price?
The efficiency loss becomes a benefit — the battery is paid to consume, and inefficiency means it consumes more. Charging at −$20/MWh gives an effective cost of −$22.73, so it only needs to sell above $7.27 to be ahead.
What if more opportunities clear the threshold than the warranty allows?
Then the rule changes from "take everything above the threshold" to "take the best N," where N is the annual cycle budget. The operating threshold becomes the Nth best spread of the year, and whichever is higher — that or the economic minimum — governs.
Does the warranty affect the operating rule?
Directly. A 4,000-cycle warranty implies $45/MWh of degradation cost against $22.50 at 8,000 cycles, which roughly doubles the minimum spread and materially reduces how often cycling is worthwhile.
Closing: Price the Thing You Are Consuming
The round-trip efficiency figure is on every specification sheet, quoted to a decimal place, and it is the number dispatch models treat most carefully. It deserves rather less of that attention than it gets.
The resource a battery actually consumes when it cycles is its warranty throughput, and that consumption has a price — the pack cost divided by the cycles the warranty permits. At realistic parameters it is an order of magnitude larger than the efficiency loss, and it is the number that should be sitting in the operating rule.
Getting it into the rule changes behaviour in a specific direction: fewer cycles, taken only on wider spreads. That is a less busy asset than an efficiency-only rule would produce, and a more valuable one — because the cycles it declines were the ones that were consuming more warranty than they were earning.
The honest caveat is that the figure has real uncertainty in it, and the uncertainty is one-directional in its consequences. Treating all degradation as cycle-driven is conservative and leaves cycles unrun; treating too little as cycle-driven runs the asset into its warranty early. The split is in the warranty document, and reading it is cheaper than being wrong in either direction.
The next post takes the other half of the decision, which is harder: given that the battery is charged, when should it actually sell — and what is the stored energy worth if it waits.
Sources: Modo Energy — Battery Dispatch Model · NREL — Utility-Scale Battery Storage, Annual Technology Baseline · Modo Energy — ERCOT: What Is the Value of a Cycle for Battery Energy Storage Systems? · Ember — How Cheap Is Battery Storage?