Sizing Solar + BESS: why the answer lives in sub-hourly data

The conventional way to size a solar plant is volumetric: take annual consumption, divide by expected generation per MW, apply a target offset percentage, and quote a capacity. The arithmetic is easy and the answer is incomplete in a way that flatters the project. Annual totals say nothing about when energy arrives versus when it is needed — and in a market that settles in 15-minute blocks, when is most of the answer.

A year contains 35,040 fifteen-minute blocks. A solar plant delivers its energy into perhaps a third of them, concentrated in a midday window. An industrial load spreads across all of them — shaped by shifts, processes, seasons, and holidays. The economics of a Solar + BESS system are decided block by block, in the gap between those two shapes.

The Shapes Never Match

Plot a typical industrial load curve against a solar generation curve and the mismatch is immediate. Solar rises steeply mid-morning, peaks around noon, and is gone by early evening. The load may be flat around the clock (continuous process plants), day-heavy with an evening tail (discrete manufacturing), or spiky and shift-driven. Even a good match leaves structural gaps: the early-morning ramp before sunrise, the evening shoulder after sundown, and monsoon weeks when generation halves.

The Mismatch Is Where the Money Is

Every block of mismatch has a price. Surplus solar in a midday block is banked at a haircut, compensated at a low rate, or lost — depending on the state's rules. Deficit in an evening block is bought from the grid at the peak time-of-day tariff, or exposes the consumer to deviation charges if scheduled poorly. Summed across 35,040 blocks, the mismatch cost can quietly claim a large share of the project's headline savings. Sizing that ignores the mismatch does not remove the cost; it discovers it after commissioning.

A battery earns its keep by moving energy between blocks. That single physical capability serves four economically distinct roles:

• Peak shaving: capping the demand peaks that set contract demand and billing demand charges — a few hundred well-chosen blocks a year can carry a disproportionate share of the fixed-charge bill
• Time-of-day arbitrage: charging from surplus solar (or cheap off-peak grid power) and discharging into peak-tariff evening blocks — converting the day's lowest-value energy into its highest-value energy
• Settlement compliance: absorbing schedule deviations and shaping injection under 15-minute DSM regimes, so that what the plant delivers matches what was declared
• Backup: riding through grid interruptions that would otherwise trip sensitive processes — a role that competes for the same stored energy as the economic roles above

The roles overlap in time and compete for the same megawatt-hours. A battery cannot simultaneously hold reserve for backup and empty itself into the evening peak. Which role wins in which block (and therefore how much power and how much energy the battery needs) is precisely what sizing has to decide.

The industry's shortcuts — "size solar to 80% of annual consumption", "add a two-hour battery at a quarter of the solar capacity" — fail because each of the four roles implies a different power-to-energy ratio, and the right blend depends on the specific load shape, the state's tariff structure, and the settlement rules. Peak shaving wants power; arbitrage wants energy; DSM compliance wants both, at the times the other two want them. A heuristic tuned to one consumer's shape transfers poorly to another's — and a heuristic tuned to yesterday's banking rules ages badly.

There is a second, subtler failure: solar and battery sizes are not independent. A larger plant creates more midday surplus, which raises the value of storage; more storage raises the tolerable plant size. Sizing them sequentially (plant first, battery after) leaves that interaction on the table. They have to be sized together.

The defensible way to size a Solar + BESS system is to pose it as a single mathematical optimisation: choose the solar capacity, wind capacity (where relevant), battery power, and battery energy — and simultaneously the battery's charge/discharge decision in every one of the 35,040 blocks — that minimise total energy cost or maximise project returns, subject to the real constraints: the tariff structure, time-of-day slabs, banking rules, deviation settlement, contract demand, and the physics of the battery itself.

This is a large-scale linear program, and it is how EarthSync's planning engine approaches it: consumption, generation, and storage modelled at 15-minute resolution, co-optimised across the full year against the market's actual rules — every one of the 35,040 blocks in play, with the financial model inside the optimisation. The formulation is chosen to stay tractable: sizing and dispatch remain one linear program even at that scale, so the solver returns a provably optimal answer in minutes, not a best effort from an overnight run. The optimiser has no attachment to round numbers. It returns configurations a rule of thumb would not propose, because it can see the block-level interactions a heuristic cannot.

The spread a battery is sized against is unusually wide on the Indian exchange. Across June 2026 the median day on the day-ahead market travelled ₹8.91/kWh from its cheapest block to its dearest. The 08:00 to 16:00 window averaged ₹1.63 while blocks from 22:00 onward averaged six times that, and 32% of all blocks in the month cleared at or above ₹9.

A spread of that size changes what the sizing question is. When the daily range is a few rupees, storage is a refinement. When the median day travels almost nine rupees and a third of blocks sit near the ceiling, the battery decides which market the portfolio sells into, and its power rating and duration set revenue rather than trim it.

It also explains why hourly resolution is inadequate here. The spread is realised inside 15-minute blocks, and an hourly average of the evening ramp smooths away precisely the intervals where the arbitrage is earned.

A single optimal answer is the beginning of the analysis, not the end. The decision-grade question is how the optimum moves when assumptions do:

• Capex: battery costs in particular are on a steep trajectory; the configuration that is optimal at today's prices may be under-stored at next year's. Testing the optimum across a capex band reveals whether to build storage now or leave space to retrofit
• Tariff escalation: most of a project's savings arrive in later years, so the assumed grid-tariff escalation rate does heavy lifting in any IRR. Run the model across escalation scenarios rather than betting on one
• Degradation: solar output declines with age, and battery capacity fades with cycling and calendar life — an augmentation plan changes both the capex profile and the effective energy over life. A 15-minute model can carry degradation year by year instead of applying a flat derate
• Regulatory drift: banking windows, ToD slabs, and DSM provisions are all moving. The most valuable output of a sensitivity run is knowing which rule change would actually hurt — and which configuration is robust to it

Annual arithmetic sizes a project for the market as it used to be. Sub-hourly optimisation sizes it for the market as it settles — 15 minutes at a time. The gap between those two answers is real money, and it compounds for twenty-five years.