Battery energy storage (BESS) for industrial sites

The first thing to get straight, because most confused conversations about batteries start here. A BESS has two independent sizes and they are quoted in different units.

• Energy, in kWh or MWh. How much it holds. This is what sets how long it can supply a load — the “how many hours” number.
• Power, in kW or MW. How fast it can take energy in or push it out. This is what sets how large a load it can supply at all — the “how big a chunk” number.

A 1,000 kWh battery that can only discharge at 250 kW cannot shave a 600 kW peak, however full it is. A 600 kW-capable battery holding only 300 kWh can shave that peak for half an hour and then it is empty. Both are real failure modes and both come from sizing one number and assuming the other.

C-rate is the link between them

The C-rate expresses power as a multiple of capacity. At 1C, a battery can move its entire capacity in one hour; at 0.5C, in two. So a 1,000 kWh system rated 0.5C discharges at up to 500 kW and takes two hours to empty.

Charge and discharge C-rates are often different, and both matter: a battery that can only charge at 0.25C needs four hours of surplus solar to fill, which a short winter day may not provide.

A fifth number, state of charge (SoC), is not a specification but a state — how full the pack is right now. It is what a dispatch simulation tracks hour by hour, and it is the reason a battery cannot simply be modelled as an annual energy discount: what it can do at 18:00 depends entirely on what it did at 13:00.

Round-trip efficiency, worked through

RTE is frequently quoted per direction, which is a trap for anyone reading quickly. Take a system quoted at 92% per direction:

So roughly one unit in six is lost in the round trip. That loss is not a rounding error in the business case — it sets a hard floor:

1. Time-of-Day arbitrage

Charge when the tariff is cheap, discharge when it is expensive, keep the difference less the round-trip loss. On an Indian C&I Time-of-Day tariff the gap between the off-peak zone and the evening peak zone is structural and published in advance, which makes this the most predictable of the three. It is also the one that does not need solar to exist at all.

2. Shifting solar into the peak

The variant that matters most on a site with a solar plant, and usually the largest of the three. Midday surplus that would have been curtailed — worth nothing — is stored and delivered into the evening peak, where it displaces the most expensive units on the bill.

This is the effect that breaks the ceiling described in the solar guide: solar alone removes cheap daytime units and leaves the expensive evening ones untouched. Storage is what reaches them. It also changes the solar answer, because surplus stops being waste — which is why the two have to be sized together rather than one after the other.

3. Peak shaving and demand charges

A large part of a C&I bill is not energy at all. Demand charges are billed on the site’s maximum draw in the billing period — measured in kVA, charged whether that peak lasted four hours or fifteen minutes.

A battery that discharges into the site’s few highest half-hours cuts the number the demand charge is billed on. The energy involved can be trivial; the saving is not. This is the value stream that most often justifies a battery on a site with no solar, and it is what the BESS-only mode in Ingro Sims is for.

And one that rarely appears in the model

Reliability. A battery covers outages and rides through sags, which on a process line with a costly restart can dwarf every tariff saving on this page. It is real value and it belongs in the conversation — it is just site-specific enough that a sizing model should not quietly price it in.

“The battery saves 18%” is not something you can derive from annual totals. Each hour of the year is a separate decision that depends on the hour before it, so the only way to get an honest number is to walk through all 8,760 hours in order.

For each hour, the model has to resolve:

• How much the site is consuming, and which tariff zone that hour sits in.
• How much the array is generating, and whether that exceeds the load.
• Whether to charge — is there surplus, is the pack below full, is there enough C-rate headroom to absorb it this hour?
• Whether to discharge — is this an expensive hour, is there charge above the depth-of-discharge floor, and does the discharge C-rate cover the load being served?
• What the state of charge is at the end of the hour, which becomes the starting condition for the next one.

Run that for a year and the outputs are no longer estimates: units drawn from the grid in each zone, units delivered from the battery, units of solar wasted, the peak demand actually reached, and the bill that follows from all of it. Run it again for a different battery size and you can compare them honestly, because both were subject to the same 8,760 hours.

That is the machinery the sizing guide describes, run across thousands of candidate sizes.

Battery capacity has a sharper optimum than solar capacity, in both directions.

Too small

The pack empties partway through the peak window. The expensive hours it fails to reach are the most expensive hours of the day, so the shortfall is worth more per unit than everything it did deliver. On a peak-shaving case it can miss the monthly maximum entirely and save nothing on demand charges.

Too large

Capacity that never cycles earns nothing and cost the same per kWh as capacity that does. Worse, an oversized pack often cannot be filled: there is only so much surplus solar in a day, and beyond that the battery either sits partly empty or has to charge from the grid, which — as the arithmetic above shows — needs a much wider spread to be worth doing.

The two builds worth naming

Between those failure modes sit two configurations that a good study should hand you separately, because they answer different questions:

The gap between those two is often large in capex and small in savings, which is exactly the trade-off a customer should be shown rather than have decided for them.

Worth saying plainly, because a tool that always recommends a battery is a tool nobody should trust.

• A flat tariff, or a narrow spread. If the peak and off-peak rates are close, round-trip losses eat the margin before capex is even considered.
• A load that is already concentrated in the cheap hours. A plant that runs hard through the night and idles in the evening has little expensive consumption to displace.
• A site with no solar surplus and no demand-charge exposure. With nothing free to store and no peak worth shaving, only pure arbitrage is left, and pure arbitrage is the thinnest of the three streams.
• A load too spiky for the C-rate. If the peaks that drive the demand charge are larger than the battery’s discharge power, the billed maximum barely moves however much capacity is installed.

In each of these the honest recommendation is solar alone, or nothing — and the same hourly simulation that sizes a battery is what demonstrates it. One exception applies regardless of the economics: in Maharashtra, storage alongside solar above 100 kW is now a requirement rather than a choice. See the regulation guide.

• Is the efficiency quoted per direction or round trip? The difference between 92% and 92% is about fifteen percent of the energy.
• Is the capacity nameplate or usable? Depth of discharge sits between the two, and proposals quote whichever is larger.
• What C-rate, charge and discharge? Without it the kWh figure cannot tell you what peak the system can actually cover.
• Does the model degrade the pack over its life? A flat fifteen-year projection at year-one capacity is optimistic by construction.
• Which hours does it claim to discharge into, and was that simulated or assumed? An annual-average model cannot answer this, and the answer is where the entire saving comes from.
• What was the tariff, and from which order? A saving computed on last year’s zones is a saving computed on the wrong bill. See the tariff guide.