The challenge

As the world sprints to shift to renewable energy, one of the most important puzzle pieces to facilitate this transition is to decarbonize the electric grid.

“Twenty-five percent of global emissions are derived from the power grid,” says Page Crahan, General Manager for Tapestry, X’s moonshot for the electric grid. “It doesn’t serve us the way it needs to anymore.”

The grid was built in the 1800s to transmit electricity from fossil-fueled power plants to cities and towns, functioning like a one-way highway with predictable energy demands. Today, clean energy sources interact with the grid differently than traditional power plants. Millions of devices—from big solar plants and wind farms to distributed generation, storage, microgrids, and electric vehicles—push and pull energy across it at all times, managed by thousands of hardworking people who move that energy where it needs to be in order to keep the lights on for their communities.

“By making the grid visible to everyone who works with it we aim to make it greener, fairer, and more reliable for everyone who depends on it.”

“Clean energy is crucial to meeting our climate goals,” Page says. “But the rapid growth of variable wind and solar resources poses a challenge to the planning and operations of our current power grid, which works the same way as it did more than a century ago. Asking a 130-year-old piece of technology to act completely differently requires a lot of change.”

This is where Tapestry hopes to help. The team is building technology to usher in a future powered by a sustainable, affordable, and reliable grid. They’re developing a range of AI-powered tools to make the grid—and the clean energy resources connecting to it—fully visible and manageable to the people who plan, build, coordinate, and operate every aspect of the world’s electricity system.

Tapestry’s holistic platform aims to revolutionize grid operations and planning. The team is creating an end-to-end solution for transmission and distribution that includes tools to transform day-to-day operations, like managing assets and load forecasting, as well as long-term planning to ensure the grid operates smoothly for decades into the future.

“By making the grid visible to everyone who works with it,” says Page, “we aim to make it greener, fairer, and more reliable for everyone who depends on it.”

Chile: A clean energy trailblazer

Countries all over the world are grappling to manage the energy transition, and Chile has emerged as a trailblazer. The country has set an ambitious goal of converting 70 percent of its total energy consumption to renewables by 2030, and becoming fully carbon neutral by 2050.

“Chile is well positioned to set an example for other nations on how to successfully decarbonize,” says Andy Ott, Tapestry’s Chief Operating Officer, who has been collaborating on a clean energy strategy with Chilean leadership since his time as CEO of PJM, the largest transmission grid operator in the United States. “Renewable energy is widely supported by both their population and their political leaders, and the country has access to abundant wind and solar resources.”

The Atacama desert in Chile’s north receives some of the highest solar irradiance in the world, and the wind from the Patagonia region and along the Arauco Gulf in the south similarly ranks among the top wind resources on the planet. But despite these resources, Chile’s electricity system is not without its challenges: a historic drought has diminished the country’s capacity for generating hydroelectric power in recent years. Moreover, connecting wind and solar resources to population centers across Chile’s long and narrow geography makes building and maintaining reliable transmission lines more costly and time-consuming to build than typical grids around the world.

“Among the challenges we face is integrating our diverse geography, spread throughout the country with a network that spans 3,100 kilometers,” says Roger Mellado, Grid Planning Lead for Coordinador Electrico Nacional (CEN), the national agency responsible for operating Chile’s energy system. “When incorporating renewable energy resources into the electrical system, we must ensure they operate as efficiently, economically, and safely as possible.”

The energy transition is a technology transition

Chile’s audacious vision and abundant environmental resources make it a natural partner for the Tapestry team. For the past three years, Tapestry has been collaborating with CEN as they develop new grid planning and operations software to accelerate the transition to renewables.

“The energy transition is, in many ways, a technology transition,” says Diego Pardow, Chile’s Minister of Energy. “Tapestry’s tools will allow us to go from an operation that’s based on blackboards to one that is based on what’s happening in reality.”

When CEN first began collaborating with the Tapestry team in 2021, it was clear that they shared the same values. Grid planner Roger Mellado has been interested in how the energy system works since his childhood growing up in Santiago. “Electrical energy is always linked to quality of life,” he says.

Roger knows firsthand how crucial the planning process is for integrating renewable energy onto the grid. Without the ability to map out a wide range of scenarios and plan for alternative resources when the wind doesn’t blow or the sun doesn’t shine, it would be extremely challenging to facilitate a reliable, resilient transition to clean energy.

“We must examine the performance and behavior of each available resource across the country, incorporate it into a model, and propose transmission infrastructure in advance,” Roger explains. “We need Tapestry’s help to speed up that process.”

Roger and the CEN team’s expertise is similarly crucial so Tapestry can build technology that can meet their specific needs. To design its Grid Planning Tool, the Tapestry team drew on CEN’s extensive experience in grid operations and managing variable renewable resources.

“While many of Tapestry’s team members bring a deep knowledge and expertise from years spent in the energy sector, relationships like the one we have with CEN are crucial to building our moonshot,” says Aviva Shwaid, Tapestry’s Head of Partnerships. “Every jurisdiction has its own rules and regulations, and our collaborators help us to understand the nuances of what will and won’t work in their specific region.”

Grid planning for a new era

Grid planning is an essential process for ensuring reliability as the electricity system evolves. Planners are tasked with looking into the future—sometimes decades out—to ensure the grid can keep up with future energy demand. They do this by using a range of tools to run hundreds of simulations that model out different future scenarios.

The results of these simulations must then be analyzed by experienced engineers, who then often have to manually, or through cumbersome bespoke tools, search through tables of highly technical information to understand any potential issues and develop solutions. Before planners can make critical energy infrastructure decisions like decommissioning a coal plant, they must ensure the grid can operate safely and reliably with a renewable energy source instead.

While this process was sufficient for planning a system with a handful of large-scale, centralized generators that can be turned on or off as needed, the replacement of these resources with dozens of smaller-scale, highly-variable wind and solar resources is exponentially more complex—and CEN’s servers are only capable of running one simulation at a time. Due to how time-consuming and labor-intensive this process is, CEN’s grid planners must prioritize which scenarios are most critical to simulate, which limits their ability to consider potentially important edge cases.

Tapestry’s grid planning tool aims to improve this process by making the experience faster, easier, and smarter across the board. The team is building technology to enable large-scale, long-term grid simulation—up to 20 years into the future—with highly granular data that allows for more efficient planning and reliable interconnection of renewables.

“We’re creating a collaborative tool that allows planners to build different future scenarios using assumptions about where people are going to consume energy and where the renewable generators will be,” explains Tapestry Product Manager Prune Chabert. “They can then quickly analyze the results and identify the best solution to any potential reliability issues.”

Tapestry’s Grid Planning Tool brings together advanced analytical techniques, enabling planners to make decisions faster and more confidently about where and how to connect clean energy resources. This includes making it possible to ask “what if” questions about extreme weather scenarios. For example, what if there was a low wind day during a heat wave, causing a spike in power usage? Using Tapestry’s Grid Planning Tool, planners can ask similar questions without time constraints and evaluate key metrics like reliability, cost, and emissions.

“Once planners understand all of the potential issues stemming from the connection of a new clean energy resource, they can evaluate projects to mitigate those issues,” Prune continues. “That might include building a new transmission line, a new battery storage system, or other grid enhancing technology. After they evaluate all the different options, they can select which will be the best solution in terms of cost, sustainability, and reliability.”

In addition to reducing the amount of time it takes to run the simulations, the Grid Planning Tool is able to run a large number of simulations in parallel, enabling grid planners to evaluate a wide range of scenarios at a dramatically faster pace. The tool’s automatic data management feature further accelerates the process, allowing CEN to migrate from a file-based system into a centralized database, which removes the need to manipulate data manually—a process that can be slow and error-prone.

Going forward, the team is also building features that will allow planners to simulate these long-term scenarios at an hourly resolution to better predict and factor in renewable resource variability, which will lead to more accurate decisions about generation and transmission planning. Using the tool, planners will be able to create new scenarios automatically and access a comprehensive set of key metrics for planners to evaluate, including the cost, system reliability, and carbon emissions associated with each scenario.

CEN is the first national system operator to test out the Grid Planning Tool. Using Tapestry’s software, CEN’s planners have reduced the amount of time required to run simulations by 86 percent, and can now run 30 scenarios simultaneously instead of just one, freeing up countless hours of their time to focus on analysis and decision making, and the Tapestry team aims to 100x the speed of CEN’s planning process by the end of 2024.  

Using Tapestry’s Grid Planning Tool, CEN’s planners can now simultaneously run 30x more simulations. They aim to 100x their planning process by the end of 2024.

Tapestry’s detailed model visualization and automated data collation gives planners a single unified view of the grid—which has never existed before—making the grid planning process more accessible for team members and cross-department collaboration. This collaborative format will benefit stakeholders beyond CEN: In 2024, Chilean renewable energy developers will begin to use the Grid Planning Tool to improve their designs and shorten the project approval process, which can be plagued by discrepancies between all the differing grid models.

For Roger and other grid planners at CEN, Tapestry is shortening the amount of time to simulate and analyze five scenarios from a week to a single day. Ultimately, CEN and Tapestry hope that these planning efficiencies will accelerate Chile’s transition to renewable energy by as much as a decade.

“Tapestry’s tools can reduce the time it takes for every stage of our planning process,” Roger says. “With Tapestry, we can analyze more alternatives, study more possibilities for renewable generation and demand growth, and have a much broader scope of insights than what we have with existing tools.”

Renewable energy forecasting

One of the key insights that sparked the Tapestry project at X was AI’s potential to offer radical new solutions to the increasingly complex problems in our rapidly-evolving electricity system.

From long-term grid planning to hour-by-hour operations, successfully transitioning to renewables requires highly accurate weather forecasting. While grid operators have always relied on forecasts to predict factors like demand for air conditioner use, the integration of large-scale wind and solar resources that also depend on the weather requires far greater accuracy.

Recently, Tapestry collaborated with Google DeepMind’s GraphCast team to build a wind forecasting model that could be integrated into Tapestry’s offering for CEN’s grid operations. GraphCast’s state-of-the-art, global model delivers ten-day weather predictions with unprecedented accuracy.

“Tapestry’s AI applications are a great example of enhancing state-of-the art AI to solve deep technical problems in complex domains like the grid. By working with Google DeepMind and CEN, we were able to tailor GraphCast’s original model to Chilean wind farms and retrain it to focus on key variables like wind magnitude specific to their locations.” — Ravi Jain, Head of Product and Technology for Tapestry

Tapestry and GraphCast’s collaborative model not only draws on data that is measured by weather stations, but also includes a broad subset of atmospheric variables to make better predictions. These include five Earth-surface variables—including temperature, wind speed and direction, and mean sea-level pressure—and six atmospheric variables, such as specific humidity and temperature. The forecast could also be updated in real-time as it becomes aware of new weather patterns.

Working with CEN’s data, Tapestry engineers employed best-in-class machine learning models, including Google's Temporal Fusion Transformer and Time-series Dense Encoder, to improve load forecasting performance on the NYISO dataset, and added in covariates such as cloud cover and direct radiation to evaluate performance changes.

“Chile has unique topographic features, and associated challenges with weather forecasting,” explains Praneet Dutta, Research Engineer at Google DeepMind. “We experimented with fine-tuning Graphcast models for this use case, optimizing for wind magnitude in this region. Upon comparing its forecasting performance against HRES for recent historical data, we've observed promising improvements in medium-term wind forecasting.”

This was Tapestry’s first integration of an AI-based forecast into a real-world grid project, and together with Google DeepMind, the teams were able to create a weather forecasting model tailored for Chile that’s up to 15 percent more accurate than the High-Resolution Forecast (HRES) model produced by the European Centre for Medium-Range Weather Forecasts: the industry gold-standard weather simulation system.

“The work is showing tremendous promise. Bringing AI out of the lab and into Chile’s control room is an amazing opportunity for applying AI to a real-world problem for humanity’s benefit,” says Page. “We’re excited to continue working with the Google DeepMind team to find more opportunities to apply their incredible work to the grid.”

A bright future

Grid planners and operators charged with navigating the complexity of our rapidly-evolving electricity system face multiple challenges, but Tapestry’s work with CEN demonstrates how AI and other cutting-edge software applications can give them the tools they need to make this system work for people as well as the planet. And this is just the beginning of what’s possible.

“To plan for a future grid that’s decarbonized, we need to improve forecasting for variable renewable energy sources," Page says. “Grid operators, who are making sure the grid is operating reliably hour by hour, usually have access to these insights, but grid planners would also benefit from this information. Incorporating hourly operating forecasts into decades-ahead planning enables planners and operators to work together to find the most reliable, affordable, and low carbon solutions.”

“Tapestry makes us double our optimism.”

Minister Pardow says Chile’s work with Tapestry has transformed his perspective on his country’s ability to adapt to climate change and successfully integrate renewable resources into its grid. “I would say Chile is one of the few countries in the world where the energy transition is much more of an opportunity than a risk,” he says. “Tapestry makes us double our optimism.”

Should Chile succeed in accelerating its efforts to decarbonize its electricity system, longtime energy industry leaders like Andy think it could serve as a bellwether for other nations looking to follow suit.

“If Chile reaches its goal to decommission its coal plants and shift its energy mix towards renewables, they have the potential to serve as a roadmap for the rest of the world,” he says. “Their success will show what can be done at a country-level scale when it comes to transitioning to clean energy.”

Tapestry’s work doesn’t end in Chile, however. The team hopes that the technology developed in collaboration with CEN will ultimately benefit grid planners and operators around the world. The team is also working with partners in the United States, the United Kingdom, South Africa, and Australia to help optimize the management of their electricity systems.

“Climate change is scary, but that’s precisely why it’s galvanizing people in a new way,” says Page. “These moments of great urgency are also opportunities to accomplish things that were previously thought impossible.”

Data notes: Throughout this case study “30x simulations” refers to moving from one simulation at a time per server, to 30 simulations running in parallel. “86% faster” refers to going from 120 hours to do 10 hydro-thermal optimization simulations, to taking 16 hours to do the same thing (PLP runs with cold start initiation).