Solar for commercial and industrial sites in India

Two arrangements dominate commercial and industrial solar in India, and they are sized by completely different logic.

Behind the meter

The plant sits on the consumer’s own premises — rooftop, carport, or ground-mount on adjacent land — and its output is consumed before the DISCOM’s meter ever sees it. Every unit generated and used on site is a unit not bought, so it is worth the full landed tariff: energy charge, Time-of-Day adder, electricity duty, fuel adjustment and all. That is the number that makes C&I solar work, and it is typically two to three times what the same unit would fetch if exported.

Open access

The plant sits elsewhere and its output is wheeled to the site across the grid, against a separate set of charges — wheeling, cross-subsidy surcharge, additional surcharge, banking terms. Larger, more contractual, and governed by state open-access regulations that change more often than the tariff does.

Everything below is about the first case, because that is where sizing is a modelling problem rather than a procurement one.

Here is the sentence that decides most C&I solar business cases: a unit of solar you consume is worth the landed tariff; a unit you export is worth whatever your state allows, and often that is very little.

Net metering for C&I consumers is restricted or capped in most states, net billing pays a fraction of the retail rate, and a plant behind a meter with no export arrangement at all simply throws surplus away — the inverters curtail and the energy never exists. So the value of the next kilowatt of array is not the energy it generates. It is the energy it generates that the site is drawing at that moment.

This is why the sizing curve bends. Add capacity to a small array and nearly every unit lands on live load. Keep adding, and midday generation starts overshooting midday demand — first on Sundays, then on light shifts, then on ordinary weekday afternoons. Each additional megawatt earns less than the one before it, and eventually earns close to nothing.

Why Time-of-Day makes this sharper

Since Time-of-Day tariffs became universal for Indian C&I consumers, the hours solar generates in are usually the cheapest hours on the bill. Several DISCOMs define an explicit solar-hours zone and discount it. The expensive hours — the evening peak — are precisely the hours a solar plant has stopped producing.

So solar reduces the bill by removing cheap units, while the expensive units stay. That is still a large saving, and it is usually the first thing worth doing. But it caps what solar alone can achieve, and it is the entire argument for adding storage. See Time-of-Day tariffs for how the zones are built.

The table below is a modelled solar year for each state: kilowatt-hours per megawatt-peak installed, per calendar month, built from NASA POWER satellite irradiance at the state centre for 2024. It is the same table the product uses to draw a generation preview the moment a region is picked.

What the table actually tells you

The annual total is the least interesting column. Every state in this list is inside a fairly narrow band — the gap between the best and the worst is not the difference between a good project and a bad one, and it is comfortably smaller than the difference between two tariffs.

The shape is what matters, and the shape varies enormously. Punjab swings 3.9× between May and Jan; Assam moves only 1.6×. A northern site with a weak, foggy winter and a fierce May is a different sizing problem from a southern one that runs flat all year and loses its summer to the monsoon — even where the two annual totals match.

That is because a plant sized against an annual average is oversized for half the year and undersized for the other half. In the strong months it spills surplus it cannot use; in the weak months it under-delivers against the savings the proposal promised. Which month a site’s production peak lands in, relative to its consumption peak, is a real input to the answer.

In rough order of how much they move the answer:

1. The hourly load shape. A three-shift plant running flat through the night can absorb far more solar than a single-shift unit that shuts at 18:00 — even at identical annual consumption. This is the dominant input and the one usually guessed at.
2. The tariff, and what it charges when. The value of a displaced unit is the landed rate in the zone it was displaced from. Two sites with the same load curve on two different DISCOMs get different answers.
3. Export rules. Whether surplus earns retail, earns a fraction, or simply vanishes changes the cost of oversizing from “mild” to “total”.
4. The sanctioned load and the connection. A contract demand and a transformer set a ceiling, and some DISCOMs cap plant capacity as a share of sanctioned load.
5. Available area. Real, and usually the first thing quoted — but it is a constraint on the answer, not the answer. Roof area caps how large the plant can be; it says nothing about how large it should be.
6. Seasonality, from the table above. Where the generation peak sits relative to the site’s own seasonal load.

Push a solar-only design far enough and it hits a wall built from two things at once.

The first is surplus. Past a certain capacity the midday hours are saturated and additional generation has nowhere to go. The savings curve flattens; the capex curve does not.

The second is the evening peak. The most expensive units on a C&I bill are drawn in the hours after sunset, and no amount of solar capacity touches them. A site can install enough solar to cover its entire annual consumption on paper and still pay nearly the whole of its peak-zone bill.

Those two problems have one answer, which is why they are usually solved together: store the surplus that had nowhere to go, and discharge it into the hours that cost the most. That is Time-of-Day arbitrage, and it is the subject of the storage guide. In Maharashtra, for solar installations above 100 kW, storage has also stopped being optional — see the regulation guide.

• Sizing to the roof. “We can fit 2.5 MWp” is an answer to a different question. Fit is a ceiling; use is the criterion.
• Sizing to annual units. Matching yearly generation to yearly consumption assumes energy is fungible across hours. Under a Time-of-Day tariff it is emphatically not: the units are matched and the bill barely moves, because what was displaced was cheap and what remains is expensive.
• Averaging the load. Dividing monthly units by 730 hours produces a flat line, and a flat line makes any array look fully consumed. The surplus a real load curve would have revealed shows up later as a plant that underperforms its proposal.
• Using a generic yield figure. A single kWh/kWp number applied across a state ignores the seasonal shape in the table above, and it ignores that the site is not at the state centre.
• Pricing savings at the average tariff. The bill’s blended rate is not the rate solar displaces. Solar displaces the daytime zone specifically, and on most DISCOM structures that zone sits below the average — sometimes well below.

Each of these has the same root: a monthly bill hides the hour, and the hour is where the money is. Reconstructing it is the first thing a serious study does.