14. Capstone: Design Your Own Grid

You’ve spent roughly thirty hours getting here — from the difference between energy and power, through every way humanity has found to spin a magnet, through the machine that keeps it all balanced every second, through what it costs and what it emits. This module asks you to use all of it at once.

The brief

You are handed a fictional Indian state. It has a budget, a 24-hour demand curve, and a particular geography — sunnier here, windier there, riverine somewhere else. Your job: build a portfolio of power plants and storage that keeps the lights on, without bankrupting the treasury or wrecking the atmosphere.

There is no single correct answer, because — as every module from 3 to 13 has shown you — there is no technology that wins on cost, reliability and emissions all at once. Coal is cheap and steady and dirty. Solar is cheap and clean and unreliable after dark. Nuclear is clean and steady and expensive. Storage doesn’t generate anything at all; it only moves what you already have to a better hour. Every choice you make is the same trade-off this course has been teaching from Module 3 onward, just now with all of them on the table simultaneously.

Suryagarh A sun-baked desert state, flat and dry — think Rajasthan.Pavan Pradesh A long, breezy coastline that never really stops blowing — think coastal Tamil Nadu.Nadikhand A hilly, river-fed state with real hydro potential — think Himachal or the Northeast.

Budget: ₹14,000 crore · Peak demand: 1,000 MW

Outline = demand each hour; the solid bar is what you actually supplied. Where the two do not meet, the gap is filled with diagonal hatching and a small triangle appears under that hour on the axis — those are the blackout hours in the scorecard. Build enough generation and storage below to close every gap.

Coal Cheap and steady, but the dirtiest option and slow to ramp — it cannot turn down below about 40%. 0 MW

Gas (combined cycle) Fast and cleaner than coal, and it can swing from almost nothing to full power — but fuel costs more per unit. 0 MW

Hydro Clean, fast-ramping and cheap to run — but a day only holds so much water, and only if your state has the rivers. 0 MW

Nuclear Steady and nearly emission-free, but very expensive to build and it barely moves off full power. 0 MW

Solar Cheapest to build, but only generates in daylight — capacity factor ~20%. 0 MW

Wind Cheap to run, weather-dependent, blows more at night than solar shines. 0 MW

Battery storage Shifts surplus solar/wind into the evening peak — does not add net energy. 0 MW

The lights are going out

Stay inside the budget, meet 99.5% of the day's demand with no blacked-out hours, and hold emissions under 200 gCO₂/kWh to earn “Excellent”. Miss only the emissions target and it is “Reliable, but carbon-heavy”.

Capital cost

₹0 cr budget ₹14,000 cr

Reliability

0.0% target 99.5%, and no blacked-out hours

Avg. cost

₹0.00/unit

Emissions

0 gCO₂/kWh target under 200

24 of 24 hours would see real blackouts, worst shortfall 100% of demand.

Reading your own results

Three numbers matter, and they pull against each other exactly the way real Indian power planning does:

Reliability. The chart shows every one of the 24 hours. A red bar means the demand outline was not filled — a real, hourly blackout, exactly the mechanism from Module 10’s frequency game, just measured over a day instead of a minute. Notice that reliability is decided hour by hour, not on the daily average: a portfolio that generates plenty of energy across the whole day can still fail badly at 7 pm if nothing dispatchable is standing by when the sun goes down. That is the duck curve from Module 7, arriving here as a design constraint rather than a slider to play with.

Cost. Capital cost is what you pay to build it; average cost per unit is what it costs to run, blending every technology’s fuel bill (or lack of one) the way Module 12’s LCOE does. Overshoot your budget and the state simply cannot afford what you’ve designed — exactly the constraint every real DISCOM and generation company actually faces.

Emissions. Grams of CO₂ per unit delivered — the units that actually reached a customer, so clean capacity you built but curtailed earns you nothing — using the same lifecycle logic as Module 13’s table. A cheap, reliable, coal-heavy portfolio will often score well on the first two numbers and badly on this one — which is precisely India’s real situation, compressed into one screen.

What the exercise is actually testing

Not whether you can memorise a number from an earlier module. Whether you’ve absorbed the shape of the problem:

• Solar and wind are cheap per unit but need a dispatchable partner — hydro, gas, or storage — to cover the hours they can’t generate. Build a portfolio that’s all solar and watch the evening bars turn red, no matter how much of it you build.
• Storage doesn’t create energy — it launders timing. A battery sized to catch the midday surplus and release it at the 7 pm peak buys back evening hours without adding a single extra megawatt-hour of generation — in Suryagarh, 2,000 MW of solar goes from 16 blacked-out hours to 11 once you add the largest battery the budget allows. What it cannot buy back is the night. No solar-and-storage mix, in any of the three scenarios, gets out of “the lights are going out”. Storage narrows the gap the duck curve opens; something firm still has to close it.
• Coal and gas buy reliability at a real, visible cost — in money for gas, in emissions for both. There is no trick that makes either of those costs disappear; the only question is how much of each you’re willing to pay for the reliability they buy you.
• Geography changes the right answer, not just its scale. The same portfolio that comfortably wins in a sun-baked desert state can fail badly on a windy coast, and vice versa — which is exactly why Gujarat and Tamil Nadu, both serious renewable states in reality, have built genuinely different grids.

If your first attempt fails, that failure is the lesson, not a bug. Real Indian state electricity boards, national planners at the CEA, and DISCOM finance teams solve versions of this exact puzzle continuously, at a scale where a wrong answer means real blackouts and real debt — everything Module 10 and Module 12 described was this problem, played for stakes that matter.

🔥How real planners measure 'keeping the lights on'Advanced · optional

This simulation counts blacked-out hours across one stylised day, and will not award its top grade while any hour goes dark — which is already sharper than a bare percentage of demand met. Real planners need sharper still, because a single day says nothing about the question that actually matters: how often, and how badly, does a bad day arrive?

The standard measure is LOLE — Loss of Load Expectation: the expected number of hours (or days) per year when available generation cannot meet demand. Many systems worldwide plan to a standard of roughly one day in ten years, which sounds absurdly strict until you remember what a blackout costs a hospital, a factory or a data centre.

The subtler concept is capacity credit (or ELCC — Effective Load Carrying Capability): how much firm capacity is a given plant actually worth? A 100 MW gas plant is worth close to 100 MW, because it runs when told. A 100 MW solar farm is worth far less, because the annual peak in India lands on a hot summer evening after sunset.

Here is the part that surprises people, and it is the deepest lesson of this capstone: solar’s capacity credit falls as you build more of it. The first gigawatt of solar genuinely helps with afternoon air-conditioning peaks. By the tenth gigawatt, afternoons are comfortably covered and the binding constraint has moved to 8 pm — where additional solar contributes exactly nothing. Its energy value stays high; its capacity value collapses toward zero.

This is why every serious decarbonisation plan pairs solar with storage or firm capacity rather than simply building more panels, and why you found the same thing with the slider: past a certain point, adding solar stopped fixing your red bars.

🔥🔥The optimisation problem this is a toy version ofDeep · optional

What you just did by hand, planners do as a constrained optimisation — and the structure is worth seeing, because it makes the trade-offs mathematically explicit.

The objective. Minimise total system cost over the planning horizon:

min Σᵢ (CAPEXᵢ · Capᵢ) + Σᵢ Σₜ (VarCostᵢ · Genᵢ,ₜ)

where i indexes technologies, t indexes hours, Cap is capacity built and Gen is energy generated.

Subject to constraints, which are where the physics of this whole course reappears as algebra:

1. Energy balance, every hour — Module 10’s non-negotiable rule:

   Σᵢ Genᵢ,ₜ + Discharge,ₜ − Charge,ₜ = Demandₜ for all t

2. Capacity limits, respecting availability — Module 7’s capacity factor:

   Genᵢ,ₜ ≤ Capᵢ · Availabilityᵢ,ₜ where Availability is 1 for a running thermal plant but follows the weather trace for solar and wind.

3. Storage state-of-charge continuity — Module 11’s round-trip loss:

   SOCₜ = SOCₜ₋₁ + η·Chargeₜ − Dischargeₜ/η

4. Ramp limits — Module 4’s ramp rates:

   |Genᵢ,ₜ − Genᵢ,ₜ₋₁| ≤ RampRateᵢ · Capᵢ

5. Emissions or budget caps, as policy dictates.

With linear costs this is a linear program, solvable exactly for thousands of hours and dozens of technologies. Add integer decisions — a plant is built or it isn’t, a unit is on or off — and it becomes a mixed-integer program, which is what real capacity-expansion and unit-commitment models actually solve, often overnight on serious hardware.

The honest caveat, which matters more than the mathematics: the optimum is only as good as the input assumptions. Change the discount rate (Module 12’s 🔥🔥 block showed how much that favours or punishes capital-heavy technologies), change the assumed fuel price trajectory, change the weather years you sample, and the “optimal” mix can shift substantially. This is why serious planning studies publish scenario ranges rather than a single answer — and why anyone presenting one confident number for what a 2050 grid should look like is showing you an assumption, not a result.

Careers in energy, one more time

Module 13 already named the paths this subject opens — engineering, grid operations, policy, climate finance, research. If the last hour made you want to actually build one of these portfolios for a living rather than just play with a simulation of one, that instinct is worth taking seriously. India’s transition is not finished; every module in this course describes work still being done, right now, by people who started roughly where you are.

The end of the road — and the beginning of the one you’ll actually watch unfold

You started this course with one idea: almost every power plant on Earth is just a different way of spinning a magnet. You now know exactly what spins each one, why solar breaks the rule, how the whole system stays balanced every second, what it costs, what it emits, and what India has bet on to get through the next fifty years.

That’s the course. What happens next — whether the 500 GW target holds, whether storage costs keep falling the way solar’s did, whether the coal towns get a transition that actually works, whether India’s thorium bet ever pays off — isn’t written yet. You’re equipped to actually follow it now, rather than just read the headlines about it.

⚡