Can one battery serve multiple uses? Testing Singapore’s storage vision on one timeline

Put every commitment on one worksheet

Singapore’s Energy Market Authority (EMA) launched an energy-storage vision and consultation on 26 October, with responses due on 29 January 2027. It addresses compatible uses, temporary local-network bridging and aggregation, with graduated observation and control depending on size and role under consideration. This is not a final rule or project award.[S1][S2]

The task here is not to count the uses advertised for a battery. It is to add the power required at the same time, carry the resulting stored energy forward and check whether charging restores enough energy before the next call. An offer that looks workable in separate service descriptions can fail on a shared timeline. The reader’s output is a list of commitments that cannot coexist under the stated assumptions, together with the inputs still missing.

All numerical inputs below belong to one fictional SG Group example. They are not measured equipment performance or Singapore market-product specifications. Equations, units and times are provided for a manual check on paper or in the reader’s own spreadsheet. No automatic equipment import, dispatch optimisation or eligibility assessment is offered. Safety, connection permission and contract interpretation require separate work.

Keep stored energy and connection-point power separate

Assume a 10 MW AC charging limit and a separate 10 MW AC discharging limit, with battery-side energy between 8 and 36 MWh. Initial stored energy is 24 MWh. Charging and discharging efficiencies are each assumed to be 90%; these are illustrative inputs, not a product guarantee. Standby consumption, temperature, voltage, reactive power and ramping are omitted. Passing this worksheet therefore does not authorise real operation.

MW denotes instantaneous power at the AC connection, while MWh denotes energy stored on the battery side. Delivering 1 MWh at the connection removes 1/0.9 MWh from storage; taking 1 MWh from the connection adds 0.9 MWh. Applying the same multiplication to both directions reverses the discharge-loss calculation. If a proposal quotes usable capacity on an AC basis, rebuild the equations from that definition rather than silently treating it as battery-side energy.

The 8 MWh floor is an operating boundary in this example, not a reserve automatically available for sale to another service. Nor is the 28 MWh gap between the upper and lower limits always available. Starting with 24 MWh leaves 16 MWh above the floor, equivalent to 14.4 MWh delivered at the connection. The schedule must begin with the actual starting state assumed for that day, not the capacity printed on a nameplate.

Place three services on the same six-hour timeline

Service A delivers 6 MW for hours 0–2 to cover a local shortfall. Service B adds 4 MW during the first 15 minutes. Charging is possible during hours 2–4, but local network headroom is only 2 MW. Service C then requires another 6 MW for hours 4–6. A, B and C are worksheet labels, not the names of real tariff products.

This schedule tests an actual call on B, rather than standby alone. Counting a standby payment while excluding the energy used when called disconnects the revenue story from the physical plan. Equally, a day without a call need not be treated as a day with one. Select the scenario required by the contract, test feasibility within that scenario and keep the probability of that scenario as a separate question.

Two hours between A and C do not by themselves establish recovery. The length of the interval and the power available for charging are separate inputs. A 10 MW charging capability does not permit a 10 MW draw when only 2 MW of network headroom is available. Unknown initial energy, call duration or charging headroom should remain unknown rather than receive a convenient assumption. A result cannot be more reliable than its inputs.

First check the 10 MW overlap

During the first 15 minutes, A’s 6 MW and B’s 4 MW add to 10 MW. That equals the assumed limit, so the power check passes within this simplified case. A alone requires 6 MW for the remaining hour and 45 minutes. Equality leaves no margin for errors or auxiliaries. Where equipment documents distinguish gross from net output, the check must use the output actually available at the relevant connection.

If a reader’s offer requires 5 MW for B, the total becomes 11 MW. Abundant stored energy cannot repair that 1 MW power shortfall. Options include reducing B, covering part of A with another resource or adding equipment at the required location, each with its own contractual and operating conditions. An offer that fails here should not proceed unchanged into a combined revenue table.

Check also whether A and B describe the same shortfall under different labels. Two contracts do not necessarily permit the same observed delivery to be counted twice. This example assumes distinct incremental requirements of 6 and 4 MW, totalling 10 MW. Whether an obligation is incremental, or another payment for the same action, depends on measurement and settlement terms. Adding physical requirements and adding payments are different operations.

Carry the energy balance through each interval

The first 15 minutes deliver 10 × 0.25 = 2.5 MWh AC. Storage falls by 2.5/0.9, approximately 2.778 MWh, leaving 21.222 MWh. Delivering 6 MW for the next 1.75 hours supplies 10.5 MWh AC and removes approximately 11.667 MWh from storage. Energy at hour 2 is therefore approximately 9.556 MWh. Keeping each row makes it possible to test intermediate constraints rather than merely reconcile an ending total.

A uses 12 MWh AC and B uses 1 MWh AC, a total of 13 MWh, below the initial deliverable allowance of 14.4 MWh. The floor is respected, so both initial services meet the example’s power and energy constraints. The remaining 1.4 MWh is additional AC energy deliverable above the floor, not the battery’s full 9.556 MWh state. Selling the full stored amount would overlook both the protected floor and conversion losses.

Removing B midway because it is unlikely to be called changes the scenario. Such a comparison can be useful, but it belongs in a separate column. Keep called and uncalled cases distinct and identify which the contract requires. Neither maximising every possible demand under a vague claim of conservatism nor averaging away a shortfall is sufficient. Each required scenario needs its own inputs and calculation.

Carry two hours of charging into the next call

Charging during hours 2–4 draws 2 MW × 2 hours = 4 MWh AC. At 90% efficiency, 3.6 MWh remains in storage, taking the hour-4 state to approximately 13.156 MWh. Using the charger’s 10 MW rating to assume a 20 MWh intake would ignore the 2 MW local headroom. Here recovery is limited by what the connection can receive during the interval, rather than by the charger’s rating.

C needs 12 MWh AC over two hours, removing approximately 13.333 MWh from storage while leaving the 8 MWh floor intact. Its required starting state is therefore approximately 21.333 MWh. The calculated 13.156 MWh falls short by approximately 8.178 MWh. Feasibility of A and B does not establish feasibility of the complete plan. In a sequence of commitments, one service’s ending state becomes the next service’s starting condition.

Blind subtraction through hour 6 produces approximately −0.178 MWh. That is not a physically operable negative state. It is the arithmetic result of extending an infeasible request beyond the floor. Do not plot it as realised operation. Instead identify the service that must change before the boundary is reached, keeping the unconstrained request separate from the quantity deliverable within the limits.

Find the boundary before choosing what to withdraw

Immediately before C, stored energy above the floor is approximately 13.156 − 8 = 5.156 MWh, allowing about 4.64 MWh AC to be delivered. At 6 MW, that lasts approximately 0.773 hours, or 46.4 minutes. The battery cannot cover the remaining 73.6 minutes of the two-hour request on its own. This is a conditional result for the fictional case, not a generic product duration; actual derating or auxiliary consumption could change it.

Alternatively, spreading the available energy evenly across two hours permits 4.64/2 = 2.32 MW. A location requiring 6 MW would still need another 3.68 MW. Delivering 6 MW briefly and delivering a lower constant output for two hours are different services even when their total energy is equal. If continuous 6 MW is required, a 2.32 MW result is not success. The shortfall table specifies the power and duration another resource must supply.

Arithmetic does not decide which customer has priority. Withdrawing C, moving A to another resource and declining to offer B create different obligations and payment consequences. No market priority in favour of A is assumed here. The order is a timeline for energy accounting, not a ranking of customers’ rights. Each revision should identify the service reduced, its replacement provider and the parties whose agreement is needed.

Recalculate after removing B

Test the tempting claim that removing the overlapping service B solves the problem. Avoiding its 1 MWh AC delivery leaves approximately 10.667 MWh at hour 2. Adding the original 3.6 MWh recharge gives approximately 14.267 MWh before C, still about 7.067 MWh below the required 21.333 MWh. Removing B helps but does not make C feasible. The test is the resulting state, not the mere act of deleting one use.

Without B, energy deliverable to C becomes approximately (14.267 − 8) × 0.9 = 5.64 MWh. That is 1 MWh more than the original 4.64 MWh, but still 6.36 MWh below C’s requirement. A controlled comparison reveals what declining B recovers. Comparing that recovery with B’s lost contract revenue requires separate pricing terms; recovering 1 MWh does not itself establish a profit improvement.

Nor does the example prove that more uses are inherently unsafe. A and B can coexist under the stated assumptions. The problem is the remaining constraint across later commitments, not the number of services. One long service can fail, while several services can be compatible if calls and recovery fit. The example tests which combination is feasible from a specified starting state and timeline; it does not reject multi-use storage.

Solve for the charging headroom actually required

Keeping A and B requires storage to rise from approximately 9.556 MWh at hour 2 to 21.333 MWh at hour 4. The required increase is about 11.778 MWh battery-side, or 13.086 MWh AC after allowing for charging efficiency. Over two hours, that requires approximately 6.543 MW on average. The gap above the original 2 MW headroom is approximately 4.543 MW. An instruction to charge more does not create that capacity.

The 6.543 MW figure is an average, not permission for any charging profile. If headroom varies, sum intake interval by interval while checking the 10 MW charger limit and 36 MWh storage ceiling. A proposal with no early headroom and all charging deferred to the end could exceed the charger limit despite having the same average. Finer intervals expose simultaneous limits hidden by an average.

Additional charging headroom must also fit the purpose of a temporary network bridge. If recovery introduces a large new load at the same location, verify capacity during that interval. Off-peak does not automatically mean unconstrained: demand, equipment outages and other connected loads affect the available headroom. If enabling recharge is harder than covering the original shortfall, the proposed combination needs revision.

A higher starting state needs a preceding charging history

With charging headroom fixed at 2 MW, solve instead for initial energy. A, B and C deliver 25 MWh AC in total, consuming approximately 27.778 MWh battery-side. With 3.6 MWh restored in between and 8 MWh required at the end, initial energy must be at least approximately 32.178 MWh. That is below the 36 MWh ceiling, so a feasible starting value exists in this simplified schedule. It does not arise automatically from the original 24 MWh.

Starting at 36 MWh leaves approximately 21.556 MWh after A and B, 25.156 MWh after charging, and 11.822 MWh after C. The floor is respected. A change in initial energy alone changes feasibility for the same rated equipment. This also counters an automatic assumption that new batteries are always necessary. Actual permission to charge to that state, and the ability to reach it by the start time, depend on equipment conditions omitted here.

A higher starting state has a cost before A begins. It may displace earlier services and depends on preceding charging prices and headroom. Closing the assessment at six hours can make preparation look like a free initial condition. For a repeated daily offer, connect each day’s ending state to the next day’s beginning and identify when recovery occurs. Passing the first day and sustaining the offer every day are different results.

Recheck the same schedule after usable capacity falls

Now assume usable upper energy falls from 36 to 30 MWh. This is an input change, not a forecast of degradation. The required 32.178 MWh starting state exceeds the revised ceiling, so raising initial energy alone no longer solves the schedule. Provision for replacement costs in a financial model does not permit delivery beyond the physical limit during the affected period.

Starting at 30 MWh leaves approximately 15.556 MWh after A and B. Restoring the 21.333 MWh needed for C requires approximately 5.778 MWh battery-side, or 6.420 MWh AC. That means about 3.210 MW over two hours. The original 2 MW is insufficient, but a revised charging arrangement could produce another solution. A capacity reduction need not eliminate every use; recalculate which constraint binds and which input would have to change.

For maintenance that lowers power capability, changing only the energy ceiling is insufficient. Return to the 10 MW overlap in the first 15 minutes and test available power. The affected columns depend on whether an outage involves batteries, chargers, transformers or communications. Reducing capacity in proportion to equipment count can miss these connections. A maintenance plan becomes an input only when the affected component is mapped to the constraint it changes.

Unknown inputs should not become convenient pass values

If initial energy has not been verified, calculating the required 32.178 MWh does not establish that it is available. Keep required and verified values in separate cells. With no verified value, the result is unresolved rather than automatically infeasible, and should identify the evidence needed next. Lack of capability and lack of evidence for capability are different findings; do not collapse them into one red status.

Where forecasts have ranges, test which part of the range creates a shortfall and which service reaches a boundary first. A favourable central estimate alone should not make every service saleable; contractual margins and required scenarios matter. But stacking unsupported extremes to guarantee failure is no better. Each input range, and the reason for using it, should correspond to evidence.

Measurement uncertainty matters more as energy approaches the floor. After the first two services, the example leaves only about 1.556 MWh battery-side above the boundary. That remainder, calculated with zero standby consumption and instantaneous response, is not an engineering safety margin for a real project. Appropriate margins depend on equipment, operation and contract terms; this article does not prescribe a standard percentage.

Aggregation must preserve the location of the constraint

An aggregation proposal should identify the delivery location before adding power ratings. A and C cover a shortfall at a particular connection. Spare capacity at another location cannot replace them if the intervening path lacks capacity. A service balancing the wider system and a service relieving load across a particular network section have different location requirements.

For a manual review, classify each resource as being on the relevant side of the constraint, across it, or on an unverified path. That classification does not solve network power flows, but it helps prevent unverified resources from entering the available total. A map of addresses is insufficient; the task needs a connection relationship that can support the required service. Missing network evidence should remain on the additional-information list.

Aggregation can still be valuable. Resources available at different times may expand the combinations able to meet a commitment. To calculate that benefit, keep separate power, energy and recovery rows for each asset, then add only the quantities capable of serving the required location. Applying a single duration ratio to a national total hides both individual energy floors and delivery-path limitations.

Do not allocate the same charging headroom twice

Charging overlaps matter as much as discharging overlaps. If two batteries charge through the same upstream transformer, two documents each referring to 2 MW of headroom do not necessarily create 4 MW. Determine whether that 2 MW is a shared connection limit or independent allocation to each asset. Correct equipment specifications can still produce an infeasible recovery plan when common headroom is counted twice.

Add a path table alongside the asset tables. Calculate each battery’s stored energy separately, then add charging power using the same path at the same time and compare it with path headroom. If charging is sequenced, verify that the later asset recovers before its next call. Alternating resources may remove a simultaneous-power conflict while creating a timing failure, so a changed order alone does not establish feasibility.

The aggregator’s name does not answer this question. Different firms may share a path, while one firm may operate independent paths. The number of counterparties is not the number of physically independent recovery routes. A local worksheet should connect asset ID, connection, shared segment, charging interval and allocation. This is a review of the reader’s own records; no personal or confidential operating data needs to be sent to the site.

Match each numerical input to a measurement record

The evidence for 24 MWh stored and for 10 MW delivered at the connection may come from different systems. Align battery-state estimates, AC meter energy and command receipt times to a common clock. Comparing records updated at different times as if they were simultaneous can create apparent discrepancies unrelated to actual losses. Keep the unit, measurement boundary, interval and timestamp alongside each value.

Evidence of 2 MW charging headroom may require more than a record showing that a charger once drew 2 MW. A low-load day does not establish availability throughout a future recovery interval. Check the demand and outage conditions under which capacity is available, and any duration limit. Separate forecast, allocation and actual intake columns help distinguish a planning error from an execution error.

After receiving operating records, reconcile stored-energy changes with AC energy for each interval. A difference from the simplified equation is not automatically a fault or misconduct: inspect auxiliaries, measurement boundaries, timing, assumed efficiencies and rounding. Unexplained differences should not become extra energy available for the next schedule. Revise inputs only for explained effects and leave the remainder unresolved.

Connect a command to the delivery it actually produced

For B’s 15-minute call, proof of command receipt does not prove an additional 4 MW was delivered. Connect receipt, acknowledgement, output change, arrival at the requested level, termination and the resulting energy state. A delayed response changes the energy delivered compared with the rectangular profile assumed here. Evaluate the relevant contract interval rather than substituting a sent command for measured delivery.

Where commands for different uses reach one asset, the review also needs evidence of how conflicts are resolved. That is not a claim that a network command universally outranks every contract. The worksheet needs to identify who can set priority and which service is reduced by that decision. Changing the service protected changes both the recovery schedule and the settlement outcome.

Access to data, authority to alter operation and acceptance of records for settlement are different functions. The purpose here is to identify who can verify each worksheet input, rather than create another general classification of regulation. Viewing stored energy may not confer authority to promise recharge. Conversely, operational control without the required metering leaves payment eligibility unresolved.

What the recipient of a proposal should ask for next

Applied to a proposal, the example produces specific requests: the timestamp for initial energy, A and C’s required power and duration, B’s call conditions, allocated charging headroom, shared-equipment constraints, and measurement and command records. Start with evidence for the first failing or unresolved column. More detailed revenue explanations do not resolve an offer already failing the power constraint.

Even with records, ask whether one successful test supports repeated delivery. This example produces different results from starting states of 24 and 36 MWh. A full-charge test cannot simply validate ordinary operation from a lower state. Compare test conditions with the conditions under which the service is offered, then rerun the boundaries with changed inputs. The relevant question is what the test established and what it left untested, rather than its name or scale.

The article’s worksheet is neither a submission form nor a certificate. It is a reading aid for matching a proposal to missing inputs and questions. Do not fill universal observation intervals, size thresholds or service priorities without published specifications; revisit the relevant column when they are defined. A numerical result here does not grant consultation participation, working-group membership, project selection or operating permission.

Retain feasible commitments before adding revenue

No market price or availability payment is assumed in the fictional case. Arbitrary prices could assign large revenues to physically impossible services and manufacture an investment ranking. Before entering money, a revenue table should list the delivery conditions of A, B and C and show what each revised plan can offer: the original, no B, a changed starting state or changed charging headroom. Only then apply actual contract prices and costs.

Declining B may recover 1 MWh but lose an availability or call payment. Securing more headroom to protect C may involve electricity purchases, connection rights or demand charges. Which costs apply depends on the contract; none is estimated here. Compare the same period, delivery point and continuing obligation, including preparation for both alternatives. Treating only one plan’s preparatory charging as free would distort the comparison.

Physical compatibility does not necessarily permit multiple payments. Conversely, one action may create distinct benefits for which separate payments are allowed. Keep physical delivery, entitlement to invoice and cash receipts separate. Review the distinction between revenue, profit and cash flow to separate delivery, recognition and collection. The purpose is to exclude unsupported income before assessing economics, not to guarantee profitability.

Compare alternatives against the same bridging service

C’s shortfall concerns part of a service delivering 6 MW AC for two hours at a particular location. Keep that location and period fixed when comparing alternatives. Another battery, load shifting, a connection upgrade or another supply source may not all address the same shortfall. Annual generation and nameplate capacity cannot establish a like-for-like cost ranking. Specify when, for how long and how often C is required before comparing costs.

For load shifting, place the rebound demand on the same timeline as charging. Solving C’s immediate shortfall may narrow headroom for a later recovery, requiring a longer assessment period. Stopping a process can also introduce restart energy and delivery costs. A battery’s shortfall cannot simply become an equal amount of freely shiftable demand; match it to the user’s permissible times and operational constraints.

Whether the bridging need remains after a network upgrade affects the asset’s useful commercial period. A long-term revenue assumption for a temporary need requires another use and contract after that need ends. Delayed completion may extend demand for the service while also extending maintenance and capacity obligations. A construction delay is not automatically a gain for the battery business: place additional income and additional commitments on the same timeline.

The economic channel runs through usable flexibility and its cost

In SG Group’s analysis, multi-use value is more than collecting additional payments for the same asset. Compatible commitments can increase useful capability from scarce land and connections, potentially reducing the cost of separate resources or interim arrangements. Investment based on infeasible revenue combinations can instead require later purchases of replacement supply or connection capacity. The worksheet distinguishes these possible channels; it does not measure savings.

For electricity-intensive businesses, connection timing and dependable supply affect when factories or computing facilities can operate. A feasible local-storage bridge may change commissioning times or additional procurement costs. Transmission into product prices, exports or wider investment depends on electricity contracts and business constraints. The fictional 6 MW example cannot be converted directly into national growth or global energy-price effects.

A financial-market interpretation must proceed from technical delivery to payment entitlement and then capital recovery. Without actual contracts, financing costs and failure or replacement obligations, no profit estimate for a battery supplier or utility follows. Cross-border equipment and financing channels can matter without determining share-price direction. Market reactions require separately timed price evidence, including prior expectations and competing news.

Specify what would change the assessment

The original 24 MWh plan fails because stored energy before C is insufficient. Evidence that changes that judgment is therefore specific: a verified higher starting state, secured charging headroom, a changed C requirement or another resource able to cover the shortfall at the required location. A photograph of large equipment or a provider’s track record does not repair that particular gap. Recalculate when the relevant condition changes.

A plan feasible from full charge can also weaken if the floor or usable ceiling changes. Lower maintenance power, shared charging headroom or an earlier next call affect different columns. Replace a vague statement that risk has increased with the affected service and the shortfall in MW, MWh or minutes. Where evidence does not permit quantification, retain uncertainty rather than invent a failure probability.

Where contracts and operation establish that calls do not overlap, the objection to multi-use narrows. Unsupported simultaneous maxima should not be imposed as a sales condition. Different required scenarios can legitimately produce different results for the same equipment. The relevant discipline is alignment among the scenario the buyer needs, the scenario the seller promises and the scenario the calculation tests, rather than a general preference for optimism or pessimism.

Turn your own proposal into a specific next question

Start by moving every proposed use onto one timeline. Enter initial energy and its bounds, connection-point charging and discharging limits, each service’s required power and duration, and intervening charging headroom. Resolve AC-versus-battery-side definitions before combining records. There is no need to make the proposal’s numbers resemble this example. The goal is to preserve the sequence of states under the reader’s own conditions.

Then check simultaneous power, energy after each interval and the starting condition for the next call. If a service fails, state which commitment is withdrawn and recalculate the same table. Keeping before-and-after versions identifies the service lost and capability recovered. Mark missing values unresolved and identify the evidence and party able to supply them. A general request for more information can become a precise question about which connection constraint a charging allocation shares and how long it remains available.