What can a school of fish teach us about grid incentives?

Our low-voltage grids were sized for household electricity, but are filling up fast with large prosumers — a rooftop PV system, a heat pump and a wallbox on one connection, all peaking together. ewz can rarely control what happens behind the meter. But it can price it. 

This challenge investigates one specific instrument: retroactive nodal prices. Nodal, because the same kilowatt-hour does very different things to the grid depending on where and when it flows, so it is priced per connection point rather than uniformly. Retroactive, because the price for an interval is issued only once that interval is over, computed from what the network actually did. The rule is public and the resulting prices are published — nothing is hidden. But nobody can react to a price. Everybody has to anticipate it. 

The retroactive half is not a hypothetical construct: ewz.solarquartier already settles this way, determining after each interval which network tariff applies. Design such a scheme badly and you reward the wrong behaviour, or make bills unpredictable. Design it well and a few hundred self-interested households hold a feeder inside its limits with nobody coordinating them. Four ways in — take one, or take all four. One team covering everything is as welcome as four teams each going deep, and no track counts for more than another. 

• Design the price: A retroactive nodal pricing scheme that holds voltages and line loading inside limits, leaves households better off than a flat tariff, and stays defensible to whoever receives the bill. 
• Design the household: Every pricing scheme has to pass one test: can a local controller act near-optimally on a household's own measurements alone? A simple one that manages it shows the scheme is implementable; one that profits while the grid suffers has found a flaw in the price. 
• Audit it: A fairness audit across household types, or a mock bill. Who wins and who loses — the owners with a roof, or the tenants below them? Would anyone accept this? 
• Show it: A compelling view of a feeder under stress — voltages, flows and money moving through a neighbourhood across a day. The sandbox emits numbers, not pictures. Everyone works from the same reference scenario, so anything quantitative is comparable — with a flipchart scoreboard for those who want one. 

Why we think this is worth investigating: 

• No herding: A price announced in advance invites everyone to react at once, relocating the peak. A retroactive price can penalise coincidence: the congestion a crowd causes depresses the payout it was chasing. 
• No real-time infrastructure: No centralised live control path or low-latency channel into the home — settlement runs on metering data already collected after the fact. Less to operate, and a much smaller attack surface. 
• No behaviour forecast: The operator never has to estimate how much flexibility exists or when it will move, since the price is computed from what happened. The burden of anticipation shifts to household controllers — they already hold the private information involved, and their actions reveal what flexibility a household actually has and how it values it. 
• Hard-to-model physics for free: Thermal loading and the capacity it consumes downstream are priced as observed, not as predicted. 

We have not settled this ourselves. It is open research at ewz, this is the first time the sandbox leaves our building, and a team here may well produce a better mechanism than the one we brought with us. Which is why every team ships a short manifest — five minutes to fill in, and the most valuable thing you can hand back to us, including from teams whose scores go nowhere: 

• Which measurements you used — your own, or ones neighbours agreed to share — and which you wished existed 
• If you priced: how the tariff is calculated, and which real cost it tries to internalise 
• If you controlled: what your household anticipates, and how 
• If you audited: which household types you compared, and what you found that would stop this going on a real bill 
• If you visualised: what the picture made obvious that the numbers did not

Digging up a street to lay a thicker cable is slow and expensive. Every kilowatt of flexibility that appears voluntarily at the right node and the right moment is capacity that does not have to be built — and a price signal that works also decides who provides it, letting it fall to whoever can offer it at the lowest cost or inconvenience. 

This is not a distant question. Local electricity communities are forming across Switzerland right now, and inside one the internal price is not handed down by regulation — it is agreed contractually between the members or whoever operates the community. Somebody has to design that rule. 

It is being designed this year, and it will shape behaviour in these neighbourhoods for a decade. As a utility owned by the city of Zurich, we care equally about the other half of the problem. A price that works technically but punishes tenants with no roof is not a price we can deploy.

No customer data is involved. Everything you build is public, reproducible, and yours to keep. We provide our open-source Python sandbox, released at the start of the challenge — nothing to prepare or install in advance but a laptop with git and uv. 

• A low voltage consumer feeder 
• Prosumers with PV, battery and uncontrollable load 
• A solver returning grid-wide voltages and flows 
• A naive controller heuristic and a naive price signal, both working end to end — copy one, change a line, and watch the number move 
• A shared reference scenario and a scoring proxy over voltages and line loadings, so results are comparable

Bring one of these; pick up the rest on the day. 

• Python: a pricing scheme or a controller is a few dozen lines 
• Optimisation or control: MPC, heuristics, or a well-chosen greedy rule 
• Power systems: power flow, voltage limits, hosting capacity, thermal loading 
• Mechanism design or economics: incentive compatibility, market design, tariffs 
• Visualisation: a grid under stress is far more convincing when you can watch it happen 
• Energy policy & regulation: cost allocation, tariff acceptability, customer fairness 

No prior knowledge of nodal pricing is required. We are on site across both days, and the challenge is designed so that someone working on fairness and customer acceptability contributes as much as the teammate deepest in the code. 

Curious how retroactive settlement already works in practice? This explainer covers the basics of ewz.solarquartier and how a LEG works; the VSE industry standard has the actual settlement math (§6, worked examples) if you want to see it in full. Both entirely optional.