Table of Contents
The power grid is a complex system that delivers electricity from generators to consumers. Grid operators must monitor the grid constantly and make quick decisions in case of disruptions, such as storms or equipment failures. However, the current tools for grid visualization could be more user-friendly and can slow down the decision-making process.
Researchers at Pacific Northwest National Laboratory (PNNL) have created a new AI tool called ChatGrid to tackle this issue. It is a user-friendly and interactive data interaction tool that utilizes the Department of Energy’s exascale computing efforts. It provides a fresh and effortless experience for users.
How ChatGrid works
ChatGrid is a tool that takes inspiration from the rise of question-and-answer generative AI tools like GPT-4 and Bing Chat. It enables grid operators to ask questions about the grid using natural language and receive visual answers instantly. For instance, a user can inquire about the generation capacity of the top five wind power generators in the Western Interconnection, and it will generate a visualization displaying the answer.
Additionally, users have the option to personalize the visualization by including various information layers such as generation capacity, voltage, power flow, and more.
ChatGrid is powered by a publicly available large language model (LLM), which is trained on massive amounts of text from various sources, such as websites, books, newspaper articles, and scientific articles. The LLM learns the patterns and relationships between words and can generate relevant responses to questions or commands.
ChatGrid uses the LLM to produce a structured query language (SQL) that can search an internal database of grid infrastructure data. The database contains information such as the capacity and location of the power plants, but not the actual data, which is highly sensitive and protected. It can create grid visualizations and ensure the security of the nation’s grid data.
How ChatGrid helps grid operators
ChatGrid is designed to help grid operators simplify their experience and make better decisions as they monitor the grid in real-time. ChatGrid can distill vast amounts of information for easy consumption and provide visual feedback that is easy to interpret. It can also handle complex and diverse questions that grid operators may have, such as “How much power is flowing from California to Nevada?” or “What is the impact of a solar outage on the grid reliability?”
It uses visualizations that come from data synthesized by the Exascale Grid Optimization (ExaGO) model. This model was created by PNNL, four other national labs, and Stanford University. ExaGO can simulate the nation’s power grid in real-time, helping grid planners understand the impacts of disruptions. Last year, ExaGO ran for the first time on Oak Ridge National Laboratory’s Frontier supercomputer, which can do over a billion billion computations per second.
The researchers hope to get feedback from grid operators and improve ChatGrid so that it can be used in their control rooms with real-life data. They also hope to make it more accessible and easy to download from GitHub. They encourage users to experiment with different prompts and questions to get better answers from ChatGrid.
“We’re envisioning a new way to look at data through questions,” said Shrirang Abhyankar, an optimization and grid modeling researcher at PNNL who co-created ChatGrid with former PNNL intern Sichen Jin. This technology has the potential to broaden the range of questions that can be asked to a generative AI tool and improve the way we adjust the questions to get the best answers.
ChatGrid is part of the Department of Energy’s ExaScale Computing Project, funded by the Department of Energy’s Office of Science and the National Nuclear Security Administration. PNNL’s Center for AI is making progress in advancing AI technology.