Electricity Maps Blog
February 14, 2025
-
5 mins
.png)
You might not realize it yet in your daily life, but electrification is ramping up at a staggering pace on a global scale. In 2024, electric vehicles are expected to represent more than half the vehicles sold in China [1][2]. At the same time, the world is increasingly powered by low-carbon electricity sources. The growth of solar (23%) and wind (10%) electricity generation in 2023 far outpaces that of fossil fuels (0.8%) [3]. In 2023, wind power became Europe’s second most prominent source of electricity, overtaking fossil gas [4].
Promises of electrified transportation, heating, and industry might have sounded like far-fetched promises in the past, but their realization is actually right around the corner.
Already now, charging point operators control enough charging capacity to match the power output of a large nuclear power plant [5]. This marks a new paradigm for the electricity grid, as consumers are now becoming actors in its management.
As electricity grids are now powered with cleaner but also more variable sources, consumers have the ability to become active participants in decarbonization. For example, in California, the emissions associated with charging an electric vehicle can differ by a factor of two depending on whether one charges during hours when the electricity supply is cleanest versus when it is most carbon-intensive.
At Electricity Maps, we foresee a world where billions of grid-connected systems optimize when and where they consume electricity to reduce costs and carbon emissions. Providing accurate grid forecasts is one of our key contributions to enable this global shift.
In the early morning of the 11th of December 2024, the winds quieted down in Northern Europe. Winds, so common in that part of the world in the midst of winter, were so calm that most wind turbines in the Northern and Baltic seas stopped producing. At the same time, the sky was shrouded with deep clouds, depriving the lands of solar irradiation, at a time when luminosity is already at its yearly low. This resulted in a situation coined as a “dunkelflaute” where grids with high renewable penetration must scrape all of their resources to meet the load. During that period, such tension on grid resources typically translates to higher electricity prices and increased grid carbon emissions .While dunkelflautes are not uncommon, they rarely last for extended periods. In December 2024, the grid tension lasted more than two days, highlighting the need for reliable multi-day forecasts of the power grids.
Imagine that you’re an EV owner with a flexible pricing electricity contract. On December 12th at 17:00, day-ahead prices reached 873 €/MWh due to the dunkelflaute, their highest level since the energy crisis that followed the invasion of Ukraine. Having your charging service operator warn you that such grid stress is expected for the next three days, which will lead to skyrocketing electricity prices, will enable you to plan around this disruption and save significant amounts of money while at the same time relieving the grid’s tension and charging your car with cleaner electricity.
This is just one of the solutions that the new Electricity Maps forecasts enable - by providing a comprehensive prediction for the future state of grids, worldwide, across multiple days (72 hours).
Electricity Maps has been providing forecasts for the power breakdown and the carbon intensity of electricity grids since 2017. However, as use cases relying on our forecasts grew in number and reach [6][7], so did our need for a highly scalable and future-proof forecasting engine.
We wanted to follow two key principles: interpretability and consistency with the rest of our offering. Flow tracing allows us to trace back the origin of electricity across all flows on the grid and is applied to all our historical and real-time data. It is a cornerstone of our offering, as it allows our electricity mix and carbon intensity data to be aware of the critical flows of electricity between regions and thus more accurate. We knew that if we could flow trace forecasted grid states, we would be able to construct an actionable and accurate prediction of the electricity grid's future state.
And that’s now at the heart of Electricity Maps’ forecasting engine. Under the hood, combinations of thousands of machine learning models are constantly interweaving learned parameters with extensive features to predict all the components of electricity grids worldwide.
To illustrate what we mean by this, we can have a simplified look at what happens on the west coast of the contiguous United States. A myriad of balancing authorities (black circles) are linked through interconnections (black lines between circles). Within each balancing authority (which corresponds to a zone in our definitions, for example, US-CAL-CISO), we operate a set of models that each predict a specific power mode (shown illustratively with solar and nuclear above). We also model the net flow of electricity being exchanged between balancing authorities.
That means that every time we generate a new set of forecasts, we effectively model how each component (power modes, exchanges, prices, etc) of all grids will evolve in the next 72 hours.
Eventually, we reconcile all these individual forecasts by feeding them to our flow tracing algorithm. Our solar model in US-CAL-CISO tells us that tomorrow at 11:00 the production will reach X MW; our nuclear model reliably predicts the production to remain a constant baseload of Y MW; while our net flow model for the interconnection with the Western Area Power Administration - Desert Southwest (US-SW-WALC) tells us that Z MW of the peak solar production will be exported out of California. Flow tracing aggregates all these inputs to build a physically coherent configuration of all interconnected grids. It thus allows Electricity Maps to most accurately predict the future origin of electricity, in each grid, globally.
Our offering is simple. For all zones where we have data, as displayed on our app, we offer 72 hours of forecasts. Quality is guaranteed for the metrics that are most useful for decarbonization: flow-traced carbon intensity and renewable energy percentage. On a 30-day rolling window, we verify in each zone that our forecasts have an average absolute error of less than 30% of the typical carbon intensity and less than 10% of the renewable percentage. Additional and higher-quality guarantees are also provided to customers to ensure these forecasts enable tangible real-world emissions reductions.
Interested in using our forecasts? You can access our API here.
The exact definition of our metrics is provided in Annex below.
[1] China’s Nov car sales rise fastest since January, subsidised trade-ins gain steam (2024) Reuters. Available at: https://www.reuters.com/business/autos-transportation/chinas-nov-car-sales-rise-fastest-since-january-subsidised-trade-ins-gain-steam-2024-12-09/ (Accessed: 14 February 2025).
[2] China EV sales statistics (2024) Road Genius. Available at: https://roadgenius.com/cars/ev/statistics/china/ (Accessed: 14 February 2025).
[3] Ember Energy (2024). Available at: https://ember-energy.org/ (Accessed: 14 February 2025)
[4] Report from the commission to the european parliament, the council, the european economic and social committee and the committee of the regions state of the energy union report 2024 (pursuant to regulation (eu)2018/1999 on the governance of the energy union and climate action). Available at: https://energy.ec.europa.eu/publications/state-energy-union-report-2024_en (Accessed: 14 February 2025)
[5] Word of mouth from some of our customers.
[6] SmartCharging, better (2024) Monta. Available at: https://monta.com/en/blog/smartcharging-better/ (Accessed: 14 February 2025)
[7] Our data centers now work harder when the sun shines and wind blows (2020) Google. Available at: https://blog.google/inside-google/infrastructure/data-centers-work-harder-sun-shines-wind-blows/ (Accessed: 14 February 2025)
.png)
.png)
Visit our free app to understand electricity grids and their CO2 emissions around the world.
