The Five Things an EMS Actually Does: Peak Shaving, Time Shifting, Diesel-Off, Smoothing and Export Limit
An energy management system earns its place on an industrial plant by doing five things: shaving the demand peak, shifting energy between time blocks, letting the diesel set stay off, smoothing the solar ramp, and holding export at or below the limit the connection allows. Everything else an EMS supplier shows you — dashboards, reports, forecasting screens — is useful, but it is not what justifies the capital.
Being strict about the list matters because each of the five produces a different shape on the load profile, needs a different control input, and tolerates a different response time. A system specified for one and expected to do another disappoints quietly for years. This guide takes them one at a time: what each changes on the meter, what the EMS must see and act on, and where the control path has to be hard-wired because software latency is not acceptable.
The five applications, in one table
| Application | What it changes on the profile | Control input it needs | Response time it must meet |
|---|---|---|---|
| Peak shaving | Clips the top off demand peaks | Live incomer kW/kVA, aligned to the billing window | Seconds, tracking the running block average |
| Time-of-day shifting | Moves energy between blocks; total kWh unchanged | Clock, tariff block table, state of charge | Minutes |
| Diesel-off operation | Removes DG running hours | Grid and DG status, load kW, step-load capability | Sub-second for the transition |
| Solar smoothing | Flattens the rate of change of solar output | Fast solar generation measurement | Sub-second to a few seconds |
| Export limitation | Caps or prevents flow back to the grid | Directional measurement at the point of supply | Hundreds of milliseconds, hard-wired |
An EMS is the layer that makes the economic decision — when to charge, when to discharge, how hard — inside limits other systems enforce. It does not own cell safety, which belongs to the battery management system, nor grid compliance at the point of supply, which belongs to the plant controller and to protection relays.
Peak shaving: holding recorded demand below a set point
Peak shaving is discharging the battery whenever plant demand threatens to rise above a chosen set point, so the demand the meter records stays below it. The battery supplies the difference; the grid sees a flat ceiling.
The subtlety is that the meter does not record instantaneous kW. It records an average over an integration window — commonly fifteen or thirty minutes in Indian industrial supply, though the window, the quantity billed and the way demand is charged are set by the tariff order in force, so read your own order and bill before fixing a set point.
That has a direct control consequence. An EMS that reacts only when instantaneous demand crosses the set point has already lost: the block average is climbing and the battery must discharge harder for the rest of the window to pull it back. A competent controller tracks the running average inside the block, projects where it will land, and starts discharging early enough that the projection comes back under the line. The mechanics are worked through in our guide to maximum demand charge calculation.
Sizing follows from the profile, not a rule of thumb. Take twelve months of fifteen-minute data, find how far above the proposed set point the plant goes and for how long, and the kW and kWh fall out of it. The common mistake we see is a battery sized in kWh for a long evening peak and then asked for a kW the converter cannot deliver, or sized in kW for a short spike and emptied halfway through a genuine peak block.
Time-of-day shifting: moving energy, not power
Time-of-day shifting is charging the battery in one tariff block and discharging in another, so the plant buys fewer units in the expensive block. Total consumption barely changes; its distribution across the day does.
This is the application most sensitive to assumptions you do not control. Block definitions, the differential between blocks and the treatment of banked solar units are set by order and revised periodically, so build the case on the structure of your current bill and then test what happens if the differential narrows.
Shifting is also the gentlest duty, usually running at a low C-rate over several hours, which is easier on cycle life than fast peak events. Where a plant wants shifting and peak shaving from one battery, the two rarely clash in timing but do compete for state of charge.
Diesel-off operation: the battery takes what the DG was there to take
Diesel-off operation means the battery and solar carry the plant load during a grid outage, or during a window when the plant would otherwise run its generator, so the DG stays shut down. This application has the most electrical content, and it is where our own work sits closest to the problem.
For the battery to hold up a plant on its own the converter has to be grid-forming, setting voltage and frequency rather than following them. A grid-following converter cannot do this, and no EMS can make it. If diesel-off is in the scope, it must be in the converter specification from the start.
Then there is step load. How large a step a diesel set accepts without unacceptable frequency dip depends on its class and governing, and a battery converter has its own step-load and overload characteristic, usually a multiple of rating for a stated number of seconds. Put the same question to both suppliers: what is the largest single motor start this system must ride through, and what do voltage and frequency do while it does?
We build diesel synchronisation and changeover panels for plants across Gujarat, and in our experience the control philosophy is where these projects go wrong, not the hardware. A plant wants the DG to stay off, so somebody writes a scheme in which the DG start command depends on an EMS decision. Our recommendation is the opposite: keep the generator's own start-on-mains-failure logic intact and hard-wired, and let the EMS inhibit it only through a path that fails to the safe state. If the EMS drops off the network, the generator must still start. The arrangement is set out in our guide to adding a BESS to an existing DG and synchronisation panel.
Solar smoothing: flattening the ramp, not adding energy
Solar smoothing is charging or discharging the battery to limit how fast net generation changes, usually expressed as a maximum percentage of plant rating per minute. Cloud passage over a rooftop array drops output steeply and recovers just as fast, and on a weak feeder or an islanded plant that ramp appears as voltage and frequency disturbance.
Smoothing needs little energy and a lot of power. The battery makes short, frequent corrections rather than deep cycles, so the duty is characterised by cycle count and C-rate rather than kWh, and where smoothing sits alongside peak shaving on one battery the cycle budget must cover both or the warranty counters run down faster than the business case assumed.
Measurement is the hard part. Seeing a ramp requires generation data fast enough to resolve it, which usually means a dedicated measurement at the solar incomer rather than a value polled from an inverter portal. That belongs in the solar ACDB and DCDB scope, specified with metering class and output type at the start.
Export limitation: the application with a hard edge
Export limitation is preventing power flowing back into the network beyond what the connection approval permits — commonly zero for a captive plant, sometimes a defined limit. Unlike the other four it is not an economic preference but a condition of connection, and the obligations attached to interconnection and safety sit under the CEA regulations and the licensee's conditions, which you should read in the version in force.
The mechanism is straightforward: measure flow at the point of supply with a directional or bidirectional meter, and as export approaches the limit, reduce generation, absorb into the battery, or both. The difficulty is not the logic. It is the speed and integrity of the path.
| Export limitation layer | What it does | Typical response | Where it lives |
|---|---|---|---|
| Control setpoint | Trims output to hold net flow below the limit | Hundreds of milliseconds to a few seconds | Plant controller, local |
| Reverse power relay | Trips the generation breaker on sustained reverse flow | Relay time setting | Protection relay at the point of supply |
| Breaker interlock | Removes generation regardless of controller state | Breaker operating time | Hard-wired, no software in the path |
| Cloud scheduler | Changes limits, schedules and reports | Seconds to minutes, not assured | Off site |
Why export limitation cannot be left to a cloud scheduler
Three independent reasons, any one of them enough.
Latency. A round trip through a public network, a broker and an application layer is measured in seconds at best, and it is not bounded. Export control needs a deterministic response, and determinism is not something an internet path offers.
Availability. Link failure is not rare on an industrial site. A function whose failure mode is uncontrolled export must not depend on a connection that may be down for an afternoon, and a scheme that defaults to its last known setpoint has defaulted to exporting.
Accountability. When a licensee asks how export is prevented, the answer has to be demonstrable on site: a directional measurement, a relay setting you can show, a trip path you can prove by test. A log file from a remote server is not that kind of evidence.
We push back when an export scheme arrives with the limitation logic drawn inside a cloud box. The scheme we are comfortable building has three layers: a fast local controller that trims generation, a reverse power or directional relay that trips the generation breaker if the trim fails, and a hard-wired interlock passing through no processor at all. The cloud may set the target and collect the data. It does not get to be the only thing between the plant and the grid.
The measurement chain every application depends on
An EMS is only as good as what it can see, and the five applications have five different measurement requirements.
| Application | Measurement point | What it must deliver |
|---|---|---|
| Peak shaving | Main incomer | kW and kVA, fast enough to track inside the billing window |
| Time shifting | Incomer plus state of charge | Energy totals by time block |
| Diesel-off | Incomer, DG terminals, essential bus | Status, kW, voltage and frequency during transition |
| Solar smoothing | Solar incomer | Fast generation value, not an inverter portal |
| Export limitation | Point of supply | Directional measurement with a protection-class path |
Protection-class and measurement-class current transformers are not the same device, and one CT asked to do both usually does one of them badly. Specify class, burden and secondary arrangement in the panel schedule.
This is where most retrofits stall. The plant has an incomer meter producing a reading every few minutes over a serial link, which is adequate for billing reconciliation and useless for peak-shaving control. The common mistake is assuming it can be reused because it is already there. On retrofit work we have added measurement cores and separate control-grade metering far more often than we have found an installation ready to use.
When two applications want the battery at the same time
Conflicts are inevitable, and the priority stack has to be written down before commissioning, not negotiated during it. A defensible order puts safety first and economics last: battery management system limits override everything; export limitation next, because it is a connection obligation; diesel-off and outage support after that, because they protect production; peak shaving next; time shifting last, because it is the only purely economic function and the only one that can be deferred without consequence.
The consequence is a state-of-charge reserve. If the battery is to be available for an outage, a shifting schedule cannot empty it at four in the afternoon. Fix the reserve as a number in the control philosophy and test it, rather than leaving it to a supplier's default.
What the EMS must never be allowed to decide
Three things belong outside the EMS, permanently. Cell protection is one: over-voltage, under-voltage, over-temperature and over-current at cell level are the battery management system's decisions, received by the EMS as constraints it cannot argue with. Electrical protection is the second — an earth fault, a short circuit or a reverse power condition is cleared by a relay and a breaker, never by a control command. Emergency shutdown is the third, and it must be a hard-wired circuit from field device to trip coil with no processor in between.
In our FAT experience the test that reveals how a scheme was really written is the dull one: pull the communication link and watch what the plant does. A good scheme degrades to a safe, defined state — fixed output, or no output — and raises an alarm. A weak one holds its last command and carries on, which is the behaviour that causes incidents. We run that test on control panels before despatch at our works in Vadodara, and again at site. Our electrical panel FAT and routine test checklist sets out that programme.
How to prove each application at commissioning
Write an acceptance test for each application into the contract, with a pass criterion a person can observe. For peak shaving, set a temporary set point below the normal peak and show from meter data across several billing blocks that the block average stayed under it. For time shifting, show charge and discharge energy by block against the schedule. For diesel-off, open the grid breaker under realistic load, record voltage and frequency through the transition, then repeat with the largest motor starting. For smoothing, record net generation during a cloudy period against the ramp limit. For export limitation, inject a deliberate export condition twice — once with the controller active, once with it disabled — so the relay path is proved on its own.
That last point is worth insisting on. Testing a trip path with the controller still running proves only that the controller worked.
The governing principle
An EMS is an optimiser working inside limits it does not set. Give it good measurements, a written priority order and clear boundaries, and it will do all five jobs. Ask it to be the protection system as well and it will eventually fail at both.
If you are writing a specification this week, start with the measurement chain and the trip paths. They are expensive to change once the panels are built, and they decide whether the five applications are real or only drawn on a screen.
Frequently Asked Questions
What does an EMS actually do in an industrial plant?
An energy management system decides when to charge and discharge a battery and how to coordinate it with solar, the grid and a generator. In practice that is five applications: peak shaving, time-of-day shifting, diesel-off operation, solar smoothing and export limitation. It works inside limits set by the battery management system and by protection relays, and makes no safety decisions itself.
How does peak shaving reduce a demand charge?
Peak shaving discharges the battery whenever demand would otherwise rise above a chosen set point, so the demand the meter records stays below that line. Because meters record an average over an integration window rather than an instantaneous value, the controller must track the running block average and discharge before the projection crosses the line. How demand is recorded and billed is set by the tariff order in force, so read the order and your own bill.
What is the difference between peak shaving and time-of-day shifting?
Peak shaving targets power: it stops the recorded demand from exceeding a set point and is measured in kW. Time-of-day shifting targets energy: it charges during one tariff block and discharges during another, so the plant buys fewer units in the expensive block, measured in kWh. One battery can do both, but they compete for state of charge, so a written priority order is needed.
Can a battery let a plant run with its diesel generator switched off?
Yes, if the power conversion system is grid-forming, meaning it sets voltage and frequency rather than following an existing supply. The battery and converter must also accept the largest step load on the plant, typically a motor start, within acceptable voltage and frequency limits. A grid-following converter cannot support a plant on its own, whatever the EMS is told to do.
What is solar smoothing and how much battery does it need?
Solar smoothing charges or discharges the battery to limit how fast net generation changes, usually expressed as a maximum percentage of plant rating per minute. It needs little energy and high power, because corrections are short and frequent. The duty is described by cycle count and C-rate rather than kWh, and those cycles count against the battery warranty alongside any other application.
Why does export limitation need a hard-wired path?
Export limitation is a condition of connection, not an economic preference, so it must respond deterministically and keep working when communication fails. A control setpoint can trim generation quickly, but it needs backing by a reverse power or directional relay and a hard-wired interlock that removes generation regardless of controller state. A path depending on software alone cannot be demonstrated to a licensee by test.
Can a cloud platform control export limitation?
No. A round trip through a public network is measured in seconds, is not bounded, and may fail entirely for long periods, while a scheme that holds its last setpoint when the link drops has effectively defaulted to exporting. A cloud platform may set targets, change schedules and collect data, but the limitation itself belongs to a local controller, a relay and a hard-wired interlock on site.
What measurement does an EMS need at the incomer?
Peak shaving needs measurement fast enough to act within the billing window, and export limitation within a fraction of a second, neither of which a billing meter polled every few minutes can provide. Export limitation also needs a directional or bidirectional measurement at the point of supply. Protection-class and measurement-class current transformers are different devices, and both class and burden belong in the panel schedule.
What should the priority order be when applications conflict?
A defensible order is: battery management system limits first, then export limitation as a connection obligation, then outage support and diesel-off because they protect production, then peak shaving, and time-of-day shifting last as the only purely economic function. The order should also fix a state-of-charge reserve so a shifting schedule cannot leave the battery empty before an outage.
What decisions should never be moved into the EMS?
Cell-level protection belongs to the battery management system, electrical fault clearance belongs to protection relays and breakers, and emergency shutdown belongs to a hard-wired circuit running from the field device to the trip coil. The energy management system receives these as constraints and must never be able to override or delay them. A safety action that waits for a software decision is not a safety action.
How do I prove at commissioning that each EMS application works?
Write an observable pass criterion for each into the contract. Peak shaving is proved from meter data across several billing blocks against a temporary set point; diesel-off by opening the grid breaker under realistic load and recording voltage and frequency through the transition; export limitation by injecting a deliberate export condition twice, once with the controller active and once with it disabled, so the relay path is proved alone.
Related Guides
- BESS electrical panel design
- Load list and transformer/incomer sizing
- APFC panel sizing and kVAr calculation
- EMS vs BMS vs PPC vs SCADA
- Maximum demand charge calculation
- Solar banking charges and surplus
- Product page: PLC control panel
- Service: Hybrid solar + BESS
Talk to Us About Your Plant
Wisdom Techno Solutions manufactures the electrical system that sits at the centre of a solar and storage installation — the LT panel, the switchgear, the metering, the protection and the control — at our works in Vadodara, Gujarat. We build PLC control panel, and the same engineers who design an assembly test it before it ships and commission it on site.
Wisdom Techno Solutions is a licensed partner for three design-verified switchgear platforms: Rittal Ri4Power (Rittal Solutions Partner since 2021 — the Ri4Power system carries Intertek ASTA design verification to IEC 61439-1 and -2 Edition 3, with copper and aluminium busbar systems up to 6300 A), C&S CX (Licensed Partner since 2023, up to Form 4B Type II) and Siemens SIEPAN Elite/8PU (IEC 61439 Solutions Partner since 2024). We build to each system's verified configuration and routine-verify every assembly in-house at our Vadodara, Gujarat facility, for industrial, EPC and infrastructure projects across India. For a solar or storage plant that covers the LT switchboard, the BESS PCC and the AC and DC distribution.
If you want to know whether your plant is a candidate, send us twelve months of electricity bills and one day's load profile. That is enough to size a system, show where the savings would come from, and tell you honestly if the answer is no.
The quickest route is our free feasibility study: send the bills and we come back with the sizing and the numbers behind it. To work out your own numbers first, use the free time-of-day tariff calculator.
Send your single line diagram, your load list or your last twelve bills, and we will come back with the electrical scheme and the numbers behind it.
Energy management interface panels, metering and export-control schemes are built at our LT panel manufacturer in Vadodara.