Navigating the Future: AI and the Power Grid

Artificial Intelligence (AI) tools have been integrated into all types of businesses since OpenAI’s ChatGPT was announced in the fall of 2022. Since then we have witnessed an AI explosion with Google, Meta, Microsoft, and others releasing new Large Language Models, it seems, almost weekly.

For grid operators, one AI name to remember is ChatGrid, a tool being developed at the Pacific Northwest National Laboratory (PNNL) in Richland, Washington. The laboratory is the U.S. Department of Energy’s primary energy-focused research facility. Shrirang Abhyankar is an optimization and grid modelling researcher at PNNL. His focus is using AI to “simplify the experience for grid operators who have to make so many decisions as they monitor the grid in real-time.” 

At a recent World Economic Forum (WEF), experts when asked about AI tools pointed to their increasing use in current and future grid operations noting that the number of decisions utility operators will have to make “will far exceed human and conventional digital automation capabilities.” The WEF sees AI as a way to achieve an intelligent grid that doesn’t overwhelm its human operators and that is optimized to meet the needs of a decarbonized mid-21st century world.

AI tools are developing quickly but not fast enough to put the technology in charge of operating the current power grid. Human operators today use modelling tools to calculate system needs multiple times daily. These calculations factor in weather, the state of the grid and electrical distribution in different geographies, and try to anticipate where things could break. Prediction modelling can help prevent outages and minimize them when disaster strikes. This is the way the grid will be managed currently and well into the 2030s.

Introducing ChatGrid

ChatGrid is one of several AI tools designed to work with human operators of the power grid, presumably for the next decade or two. When asked about ChatGrid’s future role, Abhyankar states, “We’re envisioning a new way to look at data through questions. ChatGrid allows someone to query the data—in a literal sense—and get an instantaneous answer.”

ChatGrid has been learning using data from the Exascale  Grid Optimization (ExaGO) model developed at PNNL, Stanford University and four other U.S. laboratories. It is an enormous, columnar database containing lists of powerplants, capacity, and other essential grid components. ChatGrid reads the column labels and column contents. It uses SQL, a structured query language to search the database. Answering questions from human operators it can instantaneously produce graphs and charts. ExaGO can perform a billion billion computations per second. Even that number is still less than the calculations needed to operate the power grid, the largest machine humans have ever built.

Future versions of ChatGrid will move from the ExaGO model into the real world where the AI will observe real-time data and work with grid operators to answer their questions. ChatGrid’s end use will be the production of information actionable by human grid operators.

AI Demand and Supply Forecasting 

Electricity demand is expected to triple by the end of this decade, and then grow in multiples. The evolution of the grid will be to turn into a restructured decentralized system where power comes from a wide range of sources including traditional power plants, new nuclear fission and fusion sources, hydro, solar, wind, geothermal, kinetics, other renewables and storage. Renewables will predominate on a decarbonized grid which means weather will be a key determinant of supply.

That’s why we will see the development of AI energy forecasting tools that will look at weather, renewable energy capacity and customer demand factors to help operators optimize every resource on the power grid down to the sub-second.

AI forecasting will leverage neural networks, deep learning and machine learning to produce accurate demand forecasts. The SAS Institute, an American software company, has developed the SAS Energy Forecasting Cloud as a service for utilities and energy companies. Its forecasting tools work with AI and have been trained to use hundreds of variables. Many utilities are already using SAS AI tools. More will be joining them in the future as forecasting with renewable and distributed energy becomes more predominant.

AI and Distributed Power

Distributed power refers to decentralized sources that homes and buildings may have to obtain electricity. Today, many homes and businesses have installed rooftop solar. Building can supplement power from the grid with installed geothermal or other renewable energy sources. The evolving energy grid must account for integrating such decentralized sources of power which in a decarbonizing world will be more frequent.

Buildings producing energy and often surpluses will be active grid participants delivering locally generated power. This requires the grid to be capable of flexible loading and balancing of power distribution requiring distributed energy resource management tools or DERMs.

One DERM on the market today is the AutoGrid Flex Intelligent Energy Management software platform. It can manage millions of devices in real-time, mitigate grid, determine load balancing requirements, and optimize electrical power delivery.

AI, Transportation Electrification and the Power Grid

lenrosen4https://www.21stcentech.com

Len Rosen lives in Oakville, Ontario, Canada. He is a former management consultant who worked with high-tech and telecommunications companies. In retirement, he has returned to a childhood passion to explore advances in science and technology. More...