what would it take for a household to see, price and trade the energy it makes?
A peer-to-peer trading layer for urban rooftop solar, built on the India Energy Stack. Research, regulatory analysis, and a working seller-side module.
A rooftop owner in an Indian city generates a unit of electricity she does not use. It goes to the grid. She is paid about ₹2.50 for it. The same unit clears the market at ₹6.40 to ₹6.80. She does not know this, because nothing in her house tells her. The meter reads a net number once a month. It does not say what she produced, when she produced it, where it went, or what it was worth to whoever received it. She has made something and sold it, and she has been told neither the quantity nor the price.
- (role)
- research · systems design · python
- (period)
- 2026, ongoing
- (context)
- rooftop solar · urban indian prosumers · india energy stack
- (status)
- live research · module built, not yet piloted
What a rooftop owner is not told
Sama starts from the position that this is a legibility problem before it is a market problem. The household is already an economic actor in the grid. It just cannot see the transaction it is party to.
The export tariff and the market price are both public. The distance between them is not hidden, it is simply never put in front of the person it applies to. There is no reading, no statement, no app that resolves a month of net metering into what happened on a Tuesday afternoon in April.
An honest note on the spread. ₹2.50 against ₹6.40 is the gross gap, and it is not what a household would net. Wheeling charges, banking charges, the cross-subsidy surcharge and a platform fee all sit between the two numbers, and they vary by state and by consumer category. Establishing what actually survives those charges, under each state framework, is the first piece of work this project owes itself. A project that quotes the gross spread as the opportunity has not done that work yet.
The settlement layer arrived before the product did
India's Energy Stack went live in 2025. It applies to electricity roughly the logic UPI applied to payments: a public protocol layer that any participant can build on, rather than a private platform that has to be joined. Beckn, the protocol underneath it, describes transactions between a buyer-side platform and a seller-side platform without either of them belonging to the same company.
This is the part that changed. Peer-to-peer energy trading has been technically demonstrable for a decade and commercially impossible for almost as long, because settlement between two households required an intermediary willing to build and run it. The rails now exist as public infrastructure. What does not exist is the thing an ordinary prosumer opens on a phone.
Who is allowed to sell a unit to their neighbour
The protocol permits a trade that the state may not. Peer-to-peer trading in India is governed state by state, and the regulators are in different places: the Karnataka, Delhi and Uttar Pradesh commissions each treat open access, net metering and third-party sale under different thresholds and different surcharges. Some of this is settled, some of it is sandbox provision, and some of it is silence that a regulator has not yet had reason to break.
The reading I did across KERC, DERC and UPERC is the part of this project most likely to be wrong in a year, which is why it is written down with dates attached. The conclusion it points to is that the first deployment is not a launch but a sandbox: one feeder or one housing society, agreed with a DISCOM that has something to learn from it.
What the utility sees when a rooftop appears
Every description of peer-to-peer energy leaves out the third party at the table: the distribution company whose wires the trade runs over. Part of my work at Resilience AI was alongside distribution utility teams, on how a network could be mapped and managed with better data. It was a short engagement, and it changed the design of this project anyway.
I had thought of the grid as instrumented. It is, but unevenly. Observability is concentrated at the grid substation and thins out fast below it. The low-tension feeder, the part of the network a rooftop actually connects to, is the part the operator can see least. Peer-to-peer trading happens exactly there. That inverts the usual framing. A trading layer moving energy around a feeder the utility cannot observe is not disruptive, it is invisible, and invisible is what gets regulated away. So Sama starts from the opposite premise: the utility sees the trade, and the trade settles in terms the utility can already account for.
The second thing I took from it was about numbers rather than wires. What stopped a utility team acting on a risk score was never whether the score was accurate. It was whether anyone could say why it had come out that way. A figure with no reason attached to it is a figure nobody can defend upward to a regulator, so nothing gets spent against it. That constraint turns out to be the household's problem too, at a different scale, and it runs through the rest of this project. A number that cannot be explained does not get acted on, at any level of the system.
Sitting behind both is the structural position the sector is in, which is public and well documented rather than anything I observed. Distribution companies buy power on long-term contracts they cannot exit, carry losses between what they send out and what they bill for, and depend on commercial and industrial customers to cross-subsidise everyone else. Rooftop solar arrives into that as revenue leaving through the customers they can least afford to lose. A peer layer that only sharpens that problem has no route through a regulator. One that pays its way in visibility and dispatchable capacity might.
A seller-side module, and what it has to know
The working code is the seller-side platform. It reads generation from the inverter, holds the household's position, publishes what is available and at what price, and handles the protocol exchange through to settlement.
The hard part is not the protocol, which is specified. It is everything the module has to know before it can quote a price: what the household is about to consume, what the rooftop is about to generate under the next hour of cloud, what the household is willing to let go of, and what the grid is paying at that moment. A quote is a small forecast with a commitment attached.
What a household needs to see before it will trade
A market that a person has to reason about with a spreadsheet is a market they will not use. This is the question the project is really organised around, and it is closer to interface research than to energy engineering: what is the smallest set of things a household must see before it will choose to trade rather than default to the utility?
Some candidates, all of them testable. The live position, generating and consuming, in units rather than rupees. The price at this moment against the utility's price, as one comparison rather than two numbers. A decision that can be made once and left alone, rather than a market to be watched. And an account after the fact that says what was traded and to whom, because trust in a settlement layer is built retrospectively.
Underneath all of them is the constraint from section 03, arriving at the other end of the system. The utility would not act on a risk score nobody could explain. A household will not trade on a price it cannot reason about. Same failure, two very different rooms.
From a trade to a coordinated fleet
Peer trade is the wedge, not the business. A single household exporting three units is a rounding error to a DISCOM. Several thousand households whose inverters and batteries can be coordinated together are a dispatchable resource, which is the thing a distribution utility is short of and will pay for.
That second layer is a coordination problem across many self-interested agents with partial information, which is a research problem before it is a product. It is also the point at which this project stops being about energy markets and starts being about how a city's distributed infrastructure organises itself. And it is the version of Sama a distribution utility would want to exist, which is the only version with a route through the regulator.