Agent-based models minimizing carbon usage in flexible and efficient future integrated steelworks

Projektinformationen

AgiFlex

ID Finanzhilfevereinbarung: 101138813

DOI

10.3030/101138813

EK-Unterschriftsdatum 24 November 2023

Startdatum 1 Dezember 2023

Enddatum 30 November 2028

Finanziert unter

• Digital, Industry and Space

Gesamtkosten

€ 5 203 182,50

EU-Beitrag

€ 4 691 795,00

Koordiniert durch

ABO AKADEMI
Finland

Periodic Reporting for period 1 - AgiFlex (Agent-based models minimizing carbon usage in flexible and efficient future integrated steelworks)

Berichtszeitraum: 2023-12-01 bis 2026-05-31

Zusammenfassung vom Kontext und den Gesamtzielen des Projekts

The steel industry is undergoing a gradual transition towards more sustainable production while continuing to meet the demand for high-quality steel products. Traditional steelmaking processes remain a significant source of CO2 emissions but new production technologies, increased use of renewable electricity, and hydrogen-based processes are creating opportunities to reduce environmental impacts. However, the transition presents considerable challenges, as steel producers must make long-term operational and investment decisions under uncertain future conditions, including changes in energy availability, energy prices, carbon costs, and market requirements.

Future steelworks are expected to include combinations of established and emerging production routes. As these technologies interact within the same industrial site, managing material flows, energy systems, process gases, and emissions becomes increasingly complex. There is thus a need for decision-making tools both for planning long-term investments and operation under fluctuating market conditions during the transition to lower-carbon production.

The AgiFlex project addresses this need by developing a digital modelling and optimisation platform for integrated steelworks. The project combines digital twins of current and future production processes with agent-based optimisation methods to analyse plant-wide interactions and support decision-making. This platform can thus help identifying operating strategies and development pathways that reduce CO2 emissions, improve energy efficiency, lower fossil fuel consumption, and make more effective use of available resources, including process gases and hydrogen.

A central aspect of the developed tools is the possibility to flexibly evaluate different plant configurations rather than relying on predefined production flows. This allows exploration of different technology combinations and operating conditions, also in intermediate transition stages where conventional blast furnace-based production and newer direct reduction routes may operate simultaneously.

Overall objectives of the project are thus to develop the multi-agent framework, create digital twins for current and future steelmaking processes, analyse alternative production and energy scenarios, identify optimized decarbonisation pathways, and validate the approach through industrial demonstrations. The project also seeks to support strategic decisions concerning process integration, renewable energy use, hydrogen deployment, carbon capture technologies, and circular material flows.

Arbeit, die ab Beginn des Projekts bis zum Ende des durch den Bericht erfassten Berichtszeitraums geleistet wurde, und die wichtigsten bis dahin erzielten Ergebnisse

Initially, AgiFlex established the technical foundations of the planned platform by developing the core multi-agent system architecture, defining plausible optimisation methods, and creating the first set of digital twins representing process units in integrated steelworks.
The system architecture is a modular multi-agent framework integrating digital twins, optimisation algorithms, communication layers, and user interfaces. The architecture separates process simulation from decision-making functions and enables scalable development towards distributed optimisation and decision support. Initial versions of the framework have been implemented and tested, while centralised optimisation solutions have been established as benchmarks for future distributed approaches.

First versions of digital twins were developed for key process units including coke plants, sinter plants, blast furnaces, direct reduction shaft furnaces, basic oxygen furnaces, electric arc furnaces, secondary metallurgy, gas handling systems, power plants and hydrogen production units. Depending on data availability and process maturity, both first-principles and data-driven modelling approaches were applied. In addition, machine-learning-based forecasting models using Deep Echo State Networks were developed for renewable electricity generation and green hydrogen prices, showing promising predictive performance compared with conventional forecasting approaches.

Representative industrial scenarios for the future transitions were established for the two demonstration sites, covering intermediate stages combining blast furnace and direct reduction production routes. Where initial data was available, process unit models were adapted to units at the industrial sites, and simplified system simulations were conducted to test the models and connections.

Initial web-based human-machine interfaces were also developed, and preparations for industrial implementation have begun through validation activities using industrial datasets, establishment of remote-access environments, and procurement of demonstration hardware.

Fortschritte, die über den aktuellen Stand der Technik hinausgehen und voraussichtliche potenzielle Auswirkungen (einschließlich der bis dato erzielten sozioökonomischen Auswirkungen und weiter gefassten gesellschaftlichen Auswirkungen des Projekts)

The initial work has established the building blocks for the system in development, and main results for the project are expected as the implementation at the two demonstration sites proceeds.