Motivation

In light of the energy transition and the increasing need to reduce CO₂ emissions, industrial companies are facing growing challenges to make their energy systems more epicient, flexible, and sustainable. At the same time, the complexity of modern energy management systems continues to increase due to the multitude of available technologies, interconnected infrastructures, and dynamic data sources. For many users, it is dipicult to derive well-founded decisions from available information or to identify optimization potential. Although software?based solutions for energy management already exist, they often lack user-friendliness, intelligent data interpretation, and adaptive support in daily operations. The use of Artificial Intelligence (AI), especially LLMs, opers new possibilities: AI can analyze data automatically, provide recommendations for action, and present information in a comprehensible way. While powerful voice assistants and chatbots are already being used in other domains, their potential for industrial energy management processes remains largely untapped. The etaGPT project addresses this gap by developing an AI-supported assistant tailored to industrial requirements. This assistant will support data-based actions and contribute to accelerating the energy transition.

Objectives

The goal of the etaGPT project is to develop an interactive, AI-based assistant system for industrial energy management, built upon LLM technology. The project investigates how LLMs can be employed to analyze, assess, and optimize energy consumption data to support companies in increasing energy epiciency and reducing CO₂ emissions. A modular, application-oriented AI assistant will be developed with a user-friendly interface, adaptive interaction capabilities, and intelligent functionalities. The project aims to demonstrate both the technical and organizational integration of LLM technologies into industrial energy management systems. The result will be a robust, sustainable, and transferable system that is validated in real-world industrial settings. etaGPT thus contributes to the digitalization, decarbonization, and resilience of industrial energy systems and actively supports the implementation of the goals of the German Federal Government’s 8th Energy Research Programme.

Approach

The etaGPT project is structured into seven successive and interrelated work packages and follows a user- and application-centered development process. In the first phase, the focus is on the requirements analysis. Relevant stakeholders are identified, representative use cases are examined, and both technical and organizational requirements for the assistant system are systematically defined. Based on these findings, the second phase involves the conceptual design of the overall system. This includes the development of a scalable data architecture and the design of a system and AI architecture that ensures adaptability and robustness. In the third phase, a domain-specific LLM is developed, trained, and integrated into the system. This model is tailored specifically to energy-related and industrial tasks and is optimized to interpret domain data and support decision-making processes. The fourth phase focuses on the implementation of practical interaction functionalities, intuitive user interfaces, and seamless integration with existing energy management software. Here, the aim is to ensure usability and facilitate the adoption of the assistant in real industrial environments. In the fifth phase, the system is subjected to comprehensive testing. In addition to technical verification, particular attention is paid to evaluating the usability and user experience to ensure the assistant can be epectively operated by non-experts. This is followed by the sixth phase, which involves the practical implementation of etaGPT in selected real-world industrial applications. The system’s epectiveness, reliability, and added value are validated based on representative use cases. In the final phase, the focus is on evaluating the transferability and scalability of the developed system. This includes assessing the extent to which etaGPT can contribute to the reduction of CO₂ emissions, the improvement of energy epiciency, and the overall acceleration of the energy transition in the industrial sector.

Acknowledgment

This project is funded by the Federal Ministry for Economic Affairs and Energy (BMWE) and the project management organization Jülich. We would like to thank you for the opportunity to work on this project.

Funding source

Project sponsors

Consortium partner