How AI and Data Science Can Help Utilities Respond to the Existential Stresses on their Transmission Networks

In a recent article, CDO Magazine explored some of the pressures converging on today’s energy sector, highlighting how the transition to greener sources of energy, the explosion of AI and Large Language Models (LLMs), and the resulting growth in data centers are pushing the limits of the nation’s transmission networks. The article also offered a glimpse into how AI can also be part of the solution to this unprecedented strain on transmission networks’ operation and development.

With this article, we would like to expand on this subject and share some insights into data-driven solutions that can help energy companies meet the critical moments ahead.

AI-powered solutions for complex challenges

The ascent of AI alongside the imperative of decarbonization presents utilities with complex priorities. These include:

1. Escalating demand

2. Increasing technical debt

3. The generational erosion of institutional knowledge, as baby boomers retire

4. The integration challenges posed by variable renewable energy sources

These factors collectively heighten the risk of energy shortages and cascading failures in pivotal regions.

AI-driven solutions can offer critical help, by addressing the sector's most pressing needs: Forecasting and managing demand, precisely evaluating network integrity, and optimizing investment decisions based on robust data insights. Data Society has cultivated solutions tailored to confront these issues through collaborative engagements within the energy sector.

Enhancing maintenance efficiency and capital allocation

Many utilities measure, model, and manage transmission network health using a variety of data sources, but they often perform these tasks using aging toolkits such as Excel. Vestiges of bygone decades, these spreadsheets can be encumbered by old assumptions, broken source links, and incomplete logging of prior results.

They are often highly manual, cumbersome, and passed on — without affection — from generation to generation of employees. Bringing these tools up to date is vital for electric utilities to continue to support residential, corporate, and government consumers effectively. AI technologies imbue these legacy tools with interactivity, facilitating deeper insights and expediting decision-making processes with heightened accuracy and speed.

Each utility’s transmission’s network includes thousands of miles of high-tension overhead and underground cabling, hundreds of towers, and countless supporting components. This equipment is expensive to maintain and replace, but maintaining every mile and element of this network in properly functioning status is mandatory for servicing customers.

For example, the risk of failure associated with any particular segment of transmission line is a function of network factors (load, temperature, age, etc.) and external factors (localized weather, wind, urban versus rural, etc.).

Data Society has developed a tool that augments architecture data with near-real-time network sensor data to create an automated line risk tool that identifies and categorizes the parts of the network most at risk and incorporates learning algorithms to improve the accuracy of our predictions over time.

Technical staff at every electric utility continuously assess when it is financially beneficial to replace equipment as it nears its end of life.  There is, at any given time, significantly less funding available than a complete refresh of end-of-life equipment would demand, so complex capital allocation must be made against perceived risk.  Data Society developed an analytical tool that automates these processes, adds important new data sources to the model, makes the trade-offs tunable and transparent to all users, and looks into ways to bring a more AI-oriented, data-centric, and systematic approach to this challenge.  Appropriate solutions enable users to drill down to a detailed view of the inputs and weightings that drive top-level scores produced by the tool, quickly unlocking insights into the root causes of important or unusual findings and any adjustments needed to address them.

Improving the timeliness and precision of demand planning

As we mentioned in the prior article, certain areas of the country such as Northern Virginia and Silicon Valley have become the homes of large clusters of data centers. These regions present unique challenges for traditional demand forecasting methodologies, with the surge in data center development often outpacing the readiness of electric transmission infrastructure. It has become increasingly common for data center developers to purchase their own multi-million-dollar, transmission-scale transformers to accelerate the electrical provisioning process.

Modern data science and AI capabilities can help here too. It is possible to develop a better understanding of the demand arising from these regions by using a higher resolution analysis – for example, at a block-by-block or substation by substation level – and combining this more granular data with modern predictive analytics.

These analytical tools can offer more accurate predictions of future demand curves based on historical data and a broader portfolio of inputs. Thereafter, these granular predictions can be accumulated to give very accurate predictions of load in a hierarchical way, offering planners greater understanding of what is causing stress today and what will cause greater stress tomorrow.

Better leverage new analytics

Synchrophasor technologies, a recent addition to grid monitoring arsenals, monitor resonances on the network that are a symptom of newer sources of energy supply, and provide holistic insights akin to infrared scans in medical diagnostics. Unlike traditional Supervisory Control and Data Acquisition (SCADA) data, which offers detailed component-level information, synchrophasor data offers a broader view, locating systemic and potentially destructive frequency disturbances in regions of the network. As we discussed in the prior article, new greener sources of energy can add disturbances to the transmission network in very unpredictable ways, and synchrophasor technologies happen to be well tuned to find and describe these disturbances.

The challenge with this new technology is the scale of data to be acquired and assessed.  Synchrophasor data is high-velocity, highly granular data, with each piece of equipment generating data at least 60 samples per second and often multiples of that rate. Proper use of this new technology requires very advanced data selection, storage, and analytical techniques, and this is where AI and Machine Learning algorithms enter the picture. 

This is a new technology, and analytical techniques are still being discovered, refined, and occasionally thrown away in favor of better ones. But the combination of this new type of network health measurement, empowered by rapidly developing AI tools and techniques, makes it possible to understand fully how greener sources of energy are affecting the network, modify the network so it is more accepting of these newer sources of energy, and even increase the effectiveness and productivity of new energy sources.

Future prospects

In addition to the highlighted advancements, data science applications continue to evolve, aiding the energy sector in detecting climate risks, optimizing power distribution, and enhancing grid stability. For example, innovations such as Battery Energy Storage Systems (BESS) leverage AI algorithms to optimize energy storage and manage fluctuations in renewable energy generation.

In addition, real-time, AI-enabled monitoring through fixed and drone-based video systems promises further enhancements in network health assessment, bolstering resilience against evolving operational challenges. The many applications and uses of AI to help our transmission networks adapt to future energy supplies and demands defy enumeration.

However, new sources of energy supply – primarily solar and wind – behave in ways that are very different from traditional sources built on spinning turbines, and the network needs to evolve to better accommodate them. Across the U.S., there are 2,600 Gigawatts of new renewable sources being proposed; the entire installed base in the U.S. is now roughly 1,300 Gigawatts (up only slightly from about 1,000 Gigawatts in 2010). These new sources aren’t affecting the network, they are completely redefining what a power network is. The ongoing challenges at the intersection of AI, data science, and energy infrastructure heralds a transformative era. By harnessing these technologies, utilities can fortify their transmission networks, mitigate operational risks, and foster sustainable growth in a dynamic energy landscape.

About the Author

Fred Knops, a guest contributor to CDO Magazine, is Senior Vice President at Data Society and head of Utility Group. Data Society works with utilities across North America to help them address the transition to cleaner sources of energy, address pressing data challenges, and refine their operational process for greater speed, precision, and predictability.