Carbon Aware Pricing simulates cleaner tariffs across 38 grids

Why it matters

Dynamic rates already optimize electricity use for cost. Mekeci's project shows how carbon could become a second signal, while exposing how much the result depends on modeling choices.

Halil Yavuzhan Mekeci is running a daily experiment in electricity pricing: charge less when the grid is clean, charge more when it is dirty, and calculate how much household demand might move as a result. His Carbon Aware Pricing dashboard applies that idea to Switzerland and 37 US power regions, from California and Texas to regional utilities in Montana, Florida and the Carolinas.

Mekeci is a data engineer whose profile lists the Swiss Federal Statistical Office as his employer. He built the underlying model while completing a master's degree at the University of Geneva's Centre Universitaire d'Informatique. The university's February 2026 graduate roll places him in the master's-level digital systems and services program; the supplied research brief renders the credential as a master's degree in digital information systems.

His thesis started with a practical question: could households cut emissions by changing when they use electricity, without requiring them to consume less overall? The research aimed to test whether households could reduce emissions by moving electricity use toward cleaner hours rather than simply consuming less.

The dashboard is the product-shaped continuation of that work. A recent public listing described it as carbon-aware electricity pricing measured daily across 38 grids. Updated with new grid data each day, it compares a standard time-of-use tariff with two alternatives. Carbon-Aware Hourly adjusts prices with the grid's carbon intensity every hour. Carbon Peak Pricing keeps the normal tariff for most of the day, then applies a steep surcharge during the dirtiest hours.

A daily estimate, rather than a utility trial

For Switzerland, the dashboard's September 3rd snapshot covered 32 days from August 1st through September 1st. It estimated that hourly carbon-aware pricing would have reduced emissions by 2.4% against the standard tariff, equivalent to about 4,201 tonnes of CO2. The peak-pricing approach produced a larger estimated reduction of 3.0% over the same period.

The hourly model came out cleaner in 61% of the time blocks examined. Peak pricing helped in 44%. Those figures show why the aggregate result matters more than winning every individual interval: a pricing rule can perform poorly during some hours and still produce lower emissions over a month if it moves enough demand away from especially carbon-heavy periods.

These are counterfactual estimates. Carbon Aware Pricing applies modeled changes in electricity demand to observed grid conditions; it does not measure households receiving a deployed tariff. The figures therefore depend on how strongly consumers are assumed to respond, which appliances can move their consumption, and whether delayed demand returns later.

Mekeci's open-source thesis repository makes those assumptions unusually visible. The Swiss model sets hourly prices between half and twice the base rate as carbon intensity changes. Its critical-peak design triples the applicable time-of-use rate during the top 10% of carbon-intensive hours.

The behavioral layer uses a base price elasticity of -0.35, then adds assumptions for technology-assisted response, consumer sensitivity to price increases and a minimum demand floor intended to preserve thermal comfort. The repository also includes sensitivity and rebound-effect modules. Those choices are central to the result because the model needs to estimate how a price signal becomes a change in electricity use.

The four-year model produced a much larger result

Across Swiss grid data from 2021 through 2024, the thesis simulation reported a 19.1% emissions reduction for hourly carbon-aware pricing and an 11.0% reduction for critical-peak pricing. The hourly system changed the modeled correlation between demand and carbon intensity from positive to negative, meaning consumption moved toward cleaner periods under the simulation.

That four-year result should remain separate from the live dashboard's 2.4% estimate. The periods, inputs and grid conditions differ. The thesis uses four years of historical Swiss data and a documented behavioral model; the public dashboard reruns the pricing idea on recent data and has expanded it across dozens of US regions.

The contrast also demonstrates why a single headline percentage cannot define the concept. Carbon-aware pricing performs differently as generation mixes, weather, demand patterns and available flexibility change. Switzerland's hydro-heavy system presents a different shifting problem from ERCOT, California or PJM.

Carbon Aware Pricing says its live carbon intensity calculations use generation-mix data from ENTSO-E for Europe and the US Energy Information Administration for American regions. The thesis repository separately documents carbon-intensity data from Electricity Maps and Swiss consumption data supplied through Swissgrid. The public interface does not establish that every region uses an identical data pipeline or demand-response model.

Carbon intensity is the hard part

Mekeci's core bet is straightforward: electricity prices already tell consumers when power is expensive, so tariffs could also tell them when consumption is emissions-intensive. Smart thermostats, EV chargers, heat pumps, batteries and water heaters can react automatically, turning an hourly signal into scheduled demand rather than another chart a household has to monitor.

The difficult question is which emissions signal should drive that schedule. Carbon Aware Pricing says it calculates intensity from the generation mix, an approach that describes the average emissions associated with electricity on the grid. The generator responding to one additional kilowatt-hour can differ from that average.

WattTime's work on load shifting focuses on marginal emissions, which estimate the power plant likely to increase or decrease output when demand changes. A grid can have a cleaner average mix at a given hour while added consumption still causes a fossil-fuel generator to ramp. Transmission congestion and regional power flows complicate the calculation further.

That methodological divide does not erase the value of Mekeci's experiment. It defines the next engineering problem. A deployed tariff would need a defensible emissions signal, forecasts reliable enough to set prices ahead of consumption, and safeguards against shifting large volumes of demand into the same supposedly clean hour.

The surrounding infrastructure is already developing. Electricity Maps offers carbon-intensity, generation, load and price signals through an API, including optimizers for carbon-aware computing and smart charging. WattTime supplies marginal-emissions data for automated load shifting. Carbon Aware Pricing occupies an earlier layer: it gives researchers and tariff designers a public place to compare the rules themselves.

Mekeci has taken an academic model and made it inspectable across a broad slice of the US grid. The dashboard does not establish what a real customer would save or how a utility would recover its costs. It does make the tariff question concrete enough to test every day, as grid conditions change. For a thesis built around the timing of electricity demand, that continuing clock is the point.