As Maeve Allsup at Latitude Media pointed out - many of today’s most pressing energy trends are converging in Texas. In fact, a recent ERCOT report estimates that power demand will nearly double by 2030. This spike is a result of lots of large industries, such as AI data centers, looking for power. To meet this growing demand, Texas has abundant natural gas, solar, and wind resources, making it a focal point for the future of energy.
Several new initiatives are underway to modernize the grid, but the problem is that it takes a long time to complete them. While building new power generation facilities and transmission lines is necessary, these processes can take 10+ years to finish. None of these approaches enable both significantly expanded power and the transmission capacity needed to deliver it in the near future.
Beyond “Curtailment-enabled Headroom”
A study released by Duke University this month highlighted the “extensive untapped potential” in U.S. power plants for powering up to 100 GW of large loads “while mitigating the need for costly system upgrades.” There’s enough generating capacity to meet peak demand, the authors found, so it’s possible to add new loads as long as they’re not adding to the peak. New data centers would instead need to have limited on-site generation or storage to cover those few hours a day of missing power. This is what the authors mean by “load flexibility” and “curtailment-enabled headroom.”
As I shared with POWER Magazine last week, the Duke study illustrates that power plants do have significant untapped capacity, but the transmission grid might not. The study doesn’t address transmission constraints that can limit power delivery where it’s needed. Congestion is a real problem already without the extra load and could easily wipe out a majority of that additional capacity.
To illustrate this point, think about where you would build a large data center. Next to a nuclear plant, for example? But a nuclear plant will already operate flat out and will not have any extra capacity. The “headroom” is available on average in the whole system, not at any single power plant. A peaking gas plant might indeed be idle for most of the time, but not 99.5% of the time as highlighted by the Duke authors as the threshold. Your data center would need to take the extra capacity from a number of plants, which may be hundreds of miles apart. The transmission grid might not be able to cope with it.
Yet there is additional headroom not only in power generation, but also in the transmission grid itself. Up until recently, grid operators have not been able to maximize their grids because the technology didn’t exist to optimize line capacity with feet-level precision.
The Problem With Existing Grid Management and Static Line Ratings
Traditionally, power lines are given a static rating throughout the year, which is calculated by assuming the worst possible cooling conditions of a hot summer day with no wind. This method leads to conservative capacity estimates and does not account for environmental factors that can impact how much power can actually flow through a line.
Take the wind cooling effect, for example. Wind cools down power lines and can significantly increase the capacity of the grid. Even a slight wind blowing around four miles per hour can increase transmission line capacity by 30 percent through cooling.
That’s why dynamic line ratings (DLR) are such a useful tool for grid operators. DLR enables the assessment of individual spans of transmission lines to determine how much capacity they can carry under current conditions. DLR increases capacity by a third on average, helping utilities sell more power while bringing down energy prices for consumers.
However, DLR is not yet widely used. The core problem is that weather models are not accurate enough for grid operators. Wind is very dependent on the detailed landscape, such as forests or hills, surrounding the power line. A typical weather forecast will tell you the average conditions in the ten square miles around you, not the wind speed in the forest where the power line is. But without accurate wind data at every section, even a small portion of the line risks overheating unless the line is managed conservatively.
DLR solutions up until today have been forced to rely on sensors installed on transmission lines to collect real-time weather measurements, which are then used to estimate line ratings. But installing and maintaining hundreds of thousands of sensors is extremely time consuming, if not practically infeasible.
Screenshot from claw.gridraven.com showing the grid in Texas and the additional capacity available right now by accounting for actual weather conditions.
Gridraven’s AI Advantage to Unlocking Grid Capacity
Gridraven’s software solution can help push up to a third more power through the existing grid with DLR, right now. Our approach lies purely on data and has demonstrated safe, operable line ratings predictions several days in advance with confidence intervals that hold true.
Unlike competing DLR solutions that rely on sensors, we have created an AI-enabled weather model aimed at maximizing line capacity with pinpoint precision. The locations of every tree, building and hill in all of North America are known to within an accuracy of two feet. Gridraven incorporates this full detail into its weather forecasts and trains machine learning models based on measurements from tens of thousands of ground-based weather stations around the world.
Image of a randomly chosen substation in California, illustrating the level of detail to which all of North America has been scanned. Gridraven uses this data for precision weather forecasting.
The technology achieves two things: first, it is twice as accurate in sheltered locations as industry standard weather forecasts; second, it provides confidence intervals for each prediction. For example, when mean wind speed is 8 feet per second, the model might say with 99% confidence it is at least 5 feet per second, which would provide up to 30% more capacity.
Our software delivers extremely detailed weather forecasts to unlock one-third more grid capacity over the course of a year with no additional hardware. Unlike sensor-reliant DLR solutions that limit intelligence to real-time measurements – our technology delivers more capacity for day-ahead energy markets. To truly benefit from DLR, ratings must be known at least 48 hours ahead.
The Elering Case Study
Last year, we tested our machine learning-powered DLR system on 3,100 miles of 110 kV and 330 kV lines operated by Elering, Estonia’s TSO, predicting ratings in 15,000 individual locations. The power lines run through forests and hills, where it is not possible for conventional forecasting systems to predict conditions with precision. This DLR project was one of the largest, if not the largest, in the world. In comparison, the largest sensor roll-out in the U.S. covers only 52 measurement locations.
From September to November 2024, Gridraven’s average wind forecast accuracy was a 60% improvement over existing technology, resulting in a 40% capacity increase compared to the traditional seasonal rating. These results were further validated against actual measurements on transmission towers.
This pilot not only demonstrated the power of our solution against traditional systems, but also its reliability in challenging conditions and terrain.
You can check out the additional capacity of your grid in our live demo here.
