GridForesight

Short description

GridForesight is a project aimed at developing an agent-based decision support system to help grid operators plan and optimize grid expansion. By utilizing an agentic architecture that combines machine learning, language models, and digital twins, the tool will analyze network capacity, simulate scenarios, and propose cost-effective and sustainable solutions.

With a prototype of the GridForesight tool, which will be developed and tested in a relevant environment , electricity grid planning and development will be optimized. The tool will be able to draw alternative grid scenarios on a map based on user input, such as a new industrial establishment. The goal is to be able to deliver the right amount of electrical power to industries at the right time.

Challenges

Lack of data for training models: Although the grid companies own all their data, there is a risk of a data shortage for training the models. If this occurs, the project may have to scale down the model's capabilities.

Complexity of ML/AI: It may turn out that the ML/AI challenges required to reach the more advanced stages regarding measures for the evolving power grid and system are too extensive for the current project structure. In such a case, the mitigation would be to either scale down the model’s scope or expand the project consortium to acquire the missing expertise.

User acceptance: A risk is that users will not accept the tool or will not use it as expected. To counteract this, the project will be characterized by a design-centered work methodology where users are involved in all stages to co-develop the solutions.

Project purpose

The purpose of the project is to support power grid companies in their network expansion and development to accelerate, streamline, and optimize these processes, ensuring that the industry’s growing demand for electricity is met effectively. The project aims to investigate how a combination of traditional machine learning and large language models (LLMs) can be integrated into an AI agent solution. 

The goal is to develop a prototype of a digital tool where the agent handles time-consuming, complex workflows and simulations, while the grid planner retains control through interactive collaboration (Human-in-the-loop). This optimizes grid planning, ensuring that industries gain access to the right amount of electrical power at the right time.

Expected outcomes

• A prototype of the agent-based GridForesight tool, developed and tested in a relevant environment. This prototype will be capable of drawing up alternative electricity grid scenarios on a map based on user input, such as a new industrial establishment.
   
• An evaluation and validation of the tool's potential to optimize grid expansion and maximize the industry's power needs and connection process.
   
• A plan for scaling and disseminating the prototype, including a discussion on possible commercialization of a finished tool.
   
• The project is expected to raise the idea from TRL 4 (technology validated in a lab environment) to TRL 6 (technology demonstrated in a relevant environment).
   
• The development of new knowledge and skills in areas such as digital twins, AI/ML, industrial electrical flexibility, grid planning, and energy systems.
   
• For the participating grid companies, the project is also a vital learning experience to get started with AI/ML in their operations and to find a new solution for the significant problem of industries' need for more electricity and power.