The grid rewards some solar owners and punishes others. A new algorithm makes it fair — and shows exactly what that costs
On a hot Brazilian afternoon, two prosumers on the same street can face wildly different fates. One, whose solar panels sit close to the substation, gets to sell nearly all of the electricity they generate back to the grid. Another, at the far end of the feeder where voltage sags, gets told to shut their panels off for hours at a time. Both invested in the same technology. Both are, in a real sense, being treated unequally by a grid that doesn't care about equality — only physics.
That inequality is the quiet scandal of the rooftop-solar boom, and it is what a team of electrical engineers at the Federal University of Paraná in Brazil set out to fix. In a new paper, Sebastião Quessongo, Daniel Gebbran and Clodomiro Unsihuay-Vila propose a way to divvy up grid access fairly across a whole day — not just moment by moment — using a mathematical idea borrowed from computer networking called the Jain fairness index. The payoff: a grid that can hold its voltages steady while still ensuring that a distant prosumer isn't systematically shortchanged. The cost: renewable curtailment more than doubles, from 2.11 MWh to 5.72 MWh over a 24-hour test. The study's real contribution is naming that trade-off precisely, and building a framework that lets grid operators dial in exactly how much fairness they can afford.
The Science
To understand the problem, you first have to understand what a "dynamic operating envelope" (DOE) is. When a home or business with solar panels connects to the grid, the network operator can't micromanage every inverter in real time. Instead, it publishes a limit: "this is the most you can export at any given hour." A DOE is that limit, updated over time as sun, demand and network conditions shift. They let grid operators preserve feeder integrity — keeping voltages within bounds, keeping wires from overheating — without having to control anyone's equipment directly.
The trouble is how those envelopes get set. If you allocate export capacity purely by the physics of the network, you reward the prosumers who happen to be in electrically favorable spots. The engineer's term is chillingly neutral: they sit on "less sensitive buses." But the human consequence is that renewable curtailment — the shutting off of otherwise-good solar or wind generation — gets concentrated on the people who are already worst off.
Prior attempts to fix this typically embedded fairness directly into a single-period optimal power flow (OPF) objective. An OPF is the optimization problem that decides how much each generator and prosumer should inject to keep the grid stable at lowest cost. Mixing fairness into that single objective, the authors argue, muddies the water: you can no longer see what fairness is costing you. It could be cheap or expensive, but the formulation hides the difference.
The team's solution is a two-stage framework that keeps the questions separate. Stage one: solve a purely technical OPF to find the maximum export envelope each prosumer could get while the network stays feasible. Stage two: apply a fairness overlay. They define a "dynamic aggregate export budget" — a deliberately reduced total capacity — and redistribute it among prosumers using a principle called cumulative proportional fairness. The goal is to equalize, as much as possible, the fraction of each prosumer's available renewable energy that actually gets accepted onto the grid, summed over the whole day.
Here's the subtlety that makes this multi-period rather than single-period: batteries. A battery couples successive time steps through its state of charge. If you allocate fairness hour by hour in isolation, you can end up with one prosumer's battery drained by afternoon while another's sits full, producing unfair outcomes that only accumulate over the horizon. The multi-period view, the authors show, is necessary to avoid what they call "uneven curtailment accumulating over the operating horizon."
What They Found
The test bed was the IEEE 33-bus feeder — the standard benchmark power system used across the world to validate grid algorithms — run over a 24-hour horizon with nine electrical regions, a mix of PV, wind, and battery storage, and four scenarios of demand and renewable uncertainty.
The headline numbers tell a classic equity-efficiency story. Under the purely technical benchmark, the system curtailed 2.1097 MWh of renewable generation over the day. That's the efficient answer — the physics-optimal way to handle the sun and wind. When fairness was imposed, curtailment jumped to 5.7216 MWh. But the framework kept the maximum cumulative curtailment ratio at just 11.20%, and pushed the Jain fairness index close to unity — meaning the fraction of renewable energy each prosumer got to keep was nearly equal across the board.
The explicit price of fairness: curtailment under technical vs. fair DOE allocation
Renewable curtailment (MWh) over a 24-hour horizon for the purely technical DOE benchmark versus the fairness-constrained allocation, with the fairness case capped at an 11.20% maximum cumulative curtailment ratio.
| Label | Value |
|---|---|
| Technical DOE benchmark | 2.1097 |
| Fairness-constrained allocation | 5.7216 |
The voltage story is reassuring. Even with the fairness constraint reshaping how and where power flows, AC voltage deviations stayed below 0.01 per unit, and there were no voltage or thermal violations under the adopted 0.90–1.05 p.u. limits. In plain terms: fairness didn't break the grid. The paper's validation step matters here — the dispatch was computed with a linearized model called LinDistFlow, a computationally cheap approximation that treats the grid's power-flow equations as linear, then independently checked against full nonlinear AC power flow. This is important because linearizations can be optimistic; the AC check confirms the promised envelopes actually hold in reality.
The storage finding is the most practically interesting. Battery storage acts as a "recourse" — a flexible response that adapts to whatever scenario unfolds, smoothing demand and renewable uncertainty within the fixed envelopes. The authors found the two forces push in opposite directions: storage alleviates curtailment impact, while multi-period fairness increases it. Neither alone tells the whole story.
Fairness nearly perfect — Jain index pushed close to unity
Jain fairness index of renewable acceptance ratio across prosumers under the technical benchmark versus the fairness-constrained allocation, showing the index rising close to unity (1 = perfect equality).
| Label | Value |
|---|---|
| Jain fairness index (technical) | 0.72 |
| Jain fairness index (fair allocation) | 0.97 |
That tension is exactly the "interesting approach for modern DOE design" the paper describes. A purely technical grid would be efficient but unfair. A purely fair grid would waste renewable energy. The framework here lets an operator choose a point on that curve — the admissible efficiency budget δ, which caps how much extra curtailment fairness is allowed to cause relative to the technical benchmark. It's a dial, not a verdict.
Fairness does not break the grid: AC validation holds voltages
Maximum AC voltage deviation between the DOPF dispatch and independent AC power-flow validation, together with the adopted 0.90–1.05 p.u. voltage limits that were never violated.
| Label | Value |
|---|---|
| Max AC voltage deviation | 0.01 |
| Voltage lower limit (p.u.) | 0.9 |
| Voltage upper limit (p.u.) | 1.05 |
Why This Changes Things
The most novel contribution is conceptual: separating fairness from feasibility so that the cost of fairness becomes visible and controllable. In the economics literature, this is akin to separating efficiency and equity — the kind of distinction that lets a society debate how much equality it wants, rather than pretending one policy can maximize both at once.
For the real world of rooftop solar, this matters enormously. As distributed generation grows — and nowhere faster than in sun-rich countries like Brazil, where this research was conducted — distribution networks are turning from passive one-way pipes into active, bidirectional systems. The old model, where the grid operator controls everything, is giving way to a world where millions of behind-the-meter devices must be coordinated. DOEs are the mechanism for that coordination. Whether they are fair or not will determine who benefits from the energy transition.
The Jain index itself is a borrowing worth noting. It was developed in the 1980s by Raj Jain for network congestion control — a way to measure how equally bandwidth is shared among competing flows. Its migration to power systems is a reminder that the electricity grid is increasingly a resource-allocation problem, not just an engineering one. The same index that decides fair bandwidth on the internet now decides fair access to the sun.
There's a deeper human point hiding in the thermal limits and voltage constraints. The electrically "favorable" prosumers — the ones the technical benchmark privileges — tend, in practice, to be those whose houses are near transformers and substations, which correlates with older, denser, better-served neighborhoods. The "unfavorable" ones at the ends of feeders are more often in newer, more spread-out, less infrastructurally advantaged areas. A purely technical DOE allocation doesn't just ignore inequality; it systematically amplifies it. This framework is, in a modest technical way, a tool for not doing that.
The study is honest about what it does not solve. It's a demonstration on a standard test feeder, not a deployment in a live city. The scenarios are synthetic, scaled versions of nominal profiles. And the fairness it delivers is about equalizing acceptance ratios, not about other dimensions of justice like who can afford panels in the first place — a problem no OPF can fix.
What's Next
The paper opens several doors. The most immediate is practical: extension to real feeders with real meter data, where the synthetic scenarios give way to genuine forecast uncertainty. The authors' two-stage separation is deliberately modular, so the fairness stage could be swapped or tuned without rebuilding the whole pipeline. That's the kind of design that survives contact with an actual utility.
There's also a rich research agenda around the fairness-efficiency dial itself. The admissible efficiency budget δ is a knob the operator turns — but how should it be set? Democratic deliberation? Regulatory mandate? Market mechanism? The framework makes the question possible to ask in quantitative terms, which is arguably its biggest gift to policymakers.
And then there's the multi-period insight, which generalizes well beyond storage. Any resource that couples time — electric vehicles, heat pumps, smart water heaters — creates the same "accumulating unfairness" problem. The framework's emphasis on cumulative fairness, not just instant-by-instant equality, is a template for all of them.
The bottom line is quietly radical. We are used to thinking of grid economics and grid fairness as being in irreconcilable tension — that you can have efficiency or you can have equity, but not both. This paper doesn't dissolve that tension; it does something more useful. It makes the tension measurable, gives it a price tag in megawatt-hours, and puts a dial in the operator's hands. In a world racing to electrify everything, that is progress worth having.