How Europe is re-wiring its power grid with open source digital twins

Traditional electric utility planning and management software was never designed for the complexity required for onboarding dynamic energy systems like wind and solar, nor the more dynamic nature of data centers and electric cars. European utility operators have begun adopting PowSyBl and Power Grid Model, open-source modeling tools that might help.

Innovations in solar panels, wind turbines and batteries are rapidly outcompeting traditional gas, coal, and nuclear power plants on cost, not to mention generating less pollution and CO2. However, actually safely onboarding these into a working power grid requires navigating more complexity than traditional grid software can manage. Existing grid planning and management were built for a world of predictable, centralized control, often with one-way lanes from power plants to consumers. Traditional software tools were not designed to crunch numbers fast enough.

Peter Salemink, who leads the Power Grid Model project at Dutch distribution operator Alliander, illustrates the problem:

To give you an example of the scale, for instance, if we do low-voltage grid planning we try to plan our entire low-voltage grid for the coming years. Then we do 50 billion calculations with all kinds of potential scenarios for the coming X years ahead.

Rather than waiting for commercial vendors to catch up, European Transmission System Operators (TSOs) and Distribution System Operators (DSOs) are teaming up under the auspices of the Linux Foundation Energy (LF Energy) group to build their own electrical digital twin foundation. This approach promises greater flexibility, easier integration, and greater control than relying on large energy equipment and software vendors to get along.

Better Lego blocks

This new approach supports a Lego-block philosophy, allowing each utility to build and customize digital twins of transmission and distribution infrastructure to fit their needs.

At the transmission level, French operator RTE spearheaded work on PowSyBl (Power System Blocks – pronounced possible) for modeling the high-voltage grid connecting power plants to local distribution grids. It is an enterprise-grade Java framework designed for complex physics and strict regulatory requirements for collaborating DTOs across different regions or countries.

Nicolas Omont, Vice President of Operations at Artelys and a key contributor to PowSyBl, explains that TSOs needed a tool that could handle the intricate interplay of physics, software, and economics:

It’s hard to separate when you want to know what's going to happen. There's a physical level of a grid, then there is a software level, then there is an economic level, and all that goes together.

For example, in France, they have developed a new system to curtail wind farms in less than a minute, which is quite fast at the grid management level. This requires ensuring the windmills can be safely disconnected, the communications control system is quick enough to manage the task, and that wind farm operators are appropriately compensated. Making all this work together is a complex process requiring different kinds of expertise.

At the distribution level, Alliander has driven work on the Power Grid Model. This is the last mile of the electricity delivery system, delivering electricity to homes and businesses. This framework was written in C++ to optimize it for speed. This framework focuses on simulating the behavior of millions of distributed assets, such as air conditioners, electric vehicle chargers, heat pumps, and local battery storage. Salemink explains:

We started the project roughly five years ago because there was a need for automation and faster grid calculations to support all the needs that we have as a grid operator, and being able to connect all of our customers, which we could not do. Then we decided to build it ourselves and make it open source so we could collaborate with other parties.

Cultural aspects

The general goal of all utilities is to keep the power running and the costs down. But as it turns out, utilities in different countries tend to pursue slightly different approaches. Ormont observes that even within Europe’s interconnected system, local management philosophies dictate operational strategy.

For example, Germany is process-oriented, so the people in the control room know exactly what everyone else is going to do. This approach struggles with novel problems not planned beforehand. In contrast, in France, its is principle-based, so managers say how it could be managed overall and local experts figure out how to solve problems. This means that when the situation is new, they find a way to work through it. But the drawback can be confusion in how long it might take to solve a problem. In both countries, if you have an overload, you have twenty minutes to switch off the line. In Germany, you know the operator will do it in twenty minutes manually. In France, it's not completely true because there is uncertainty in how the front-line team will resolve the problem.

There are also different risk thresholds. In France, teams have figured out how to do live work on high-voltage lines when required. In Germany, it's not allowed for various reasons.

Meanwhile, Omont has heard a Dutch executive talk about the goal of buying and exporting power as a trading hub, but who was less concerned about whether there was enough power for the local market, since that’s the responsibility of producers. In France, power executives take it upon themselves to ensure everyone has enough power.

This cultural divide makes it impossible to impose a one-size-fits-all solution across the continent. Software has become a form of soft power, with the tools adopted by the community subtly influencing how the grid is managed across borders. Omont says:

The way the grid is managed in the Netherlands, France, and Germany is not the same. I mean, even if it's the same grid, it's not managed the same way, because it depends on how the country works. And so, especially in Europe, when we are trying to unify how we manage the grid, there is some competition over which approach will be adopted at the European level. So, there is a bit of competition. And this is where software is also a soft power, in that if you bring your software, and if the others adopt it, then it somehow helps you to push your way of doing things, except  that somehow it will be used in a way that is a bit different in each country, that will fit the organization on the culture of the country.

Need for sovereignty

European transmission and distribution operators both saw a need to take the reins of their planning and management software to gain more control, flexibility, and visibility.

Omont says that on the transmission side, operators prefer to customize their tools internally because they think they need the skills to tailor the software exactly to their needs, which enables them to make the best of the grid. They considered this doable by having the skills and the people who know how to develop the software, as well as how their grid works.

This allows them to retain power and control while sharing the development burden with other TSOs. It's similar to how telcos are starting to collaborate on open infrastructure through open radio access network (Open RAN) standards. Omont says:

TSOs are willing not to buy a big platform from GE or Siemens, but to put together building blocks the way they want them to be and to keep the control of this layer by building full applications from the smaller modules.

Salemink adds:

In the past, we would buy commercial software to do things, but it's just a black box when you don't know what happens. We would like to have more control and know exactly what happens if A goes in, then what goes out? So that's also one of the key reasons why open source is really one way to go here in the energy sector.

Planning for and with AI

AI is increasingly a compounding factor in all this, as both a massive consumer of power and a potential manager of a more complex grid. However, it's impractical to train a reinforcement learning model on the live grid, as the stakes are too high. Rather, operators would prefer to train better models across various scenarios offline and then test them extensively first.

Ormont points toward the work with Hydro Quebec on how this works:

"I’ve got a project that will start with Hydro Quebec next year, about using PowSyBl as a simulator to train foundational models for the artificial intelligence applied to the power grid that will make the simulation go faster. By using AI, we will have a gain of maybe a thousand or ten thousand on the speed for some calculations.”

My take

The shiny objects of different versions of the future, like bigger data centers, electric cars, and smarter electrified factories, tend to grab the attention. But the bottleneck to actually building all of these depends on improving digital twins' ability to plan and manage a more dynamic, scalable grid.

By treating grid modeling software as public infrastructure, much like the lines themselves, Europe is building a model that allows different cultures to collaborate and innovate at their own pace. This will be required to help clear the long connection queues, sometimes stretching into years, that will be needed to keep pace with more and bigger data centers, not to mention electric transport, factories, and heating.

One challenge in understanding power grids at a policy level is that there are so many physical, cultural, and economic factors that come together to make them work.  It got me wondering how simulation games might help. I came across Power Network Tycoon, made by David McKenzie, a utilities engineer in Australia, that helps explore the dynamics of resource constraints, safety risks, and exposure to natural disasters. I reached out to the developer, David McKenzie, who writes back:

I've been hearing a bit more about digital twins in recent years, although to be honest, I never thought of my game as being essentially a 'digital twin', although that's really what I've been going for the whole time. I love the idea of playing a city-building game, and it having enough detail that I could go out and build it in the real world, and it would function as it was simulated in-game.