Mix Matters: Technical Evaluation of a Carbon-Aware Power Plant Operation Strategy for an Industrial Area in Bremen

MIX MATTERS: TECHNICAL EVALUATION OF A CARBON-AWARE POWER PLANT OPERATION STRATEGY FOR AN INDUSTRIAL AREA IN BREMEN Danila Valko, Sharaf Alsharif and Sven Rosinger OFFIS – Institute for Information Technology Oldenburg, Germany E-mail: [email protected] KEYWORDS Carbon Footprint, Carbon Emissions, Smart Grid, Industrial Power Strategies, Co-Simulation ABSTRACT In the context of tightening climate policies, industrial clusters and large metallurgical enterprises must revise their carbon emissions targets and develop transition strategies toward carbon neutrality. This study demonstrates that reducing the carbon footprint associated with electricity production and consumption in the industrial sector can be feasible utilizing the real-time carbon intensity monitoring of on-site power plants, available distributed renewables, and green electricity supplied by the external grid. We evaluated this strategy using a co-simulation approach applied to the electrical grid of an industrial area with steel processing enterprises in Bremen.The results indicate that, given Germany’s gradual de-carbonization of electricity production, a strategy focused on adjusting on-site power generation from carbon-intensive sources such as natural gas and metallurgical waste gases can be effective in meeting stringent short-term carbon footprint targets during the energy transition. This work contributes to the open data and modeling efforts in the energy research community, ensuring reproducibility and enabling further independent research. INTRODUCTION The global imperative to reduce greenhouse gas emissions has placed significant pressure on the power sector to transition toward more sustainable energy systems. As electricity generation accounts for a large share of carbon dioxide (CO2) emissions, enhancing the carbonawareness of operational strategies within power systems is now critical for meeting climate targets (Dagnachew et al. 2021, IEA 2024). Among the various de-carbonization pathways, optimizing the operation of local generation units in response to real-time carbon intensity signals has emerged as a promising yet underexplored approach (Chen 2024). Traditional efforts to reduce power system emissions have predominantly focused on large-scale planning, such as transitioning to renewable energy sources, decommissioning fossil-fueled plants, and integrating carbon constraints into expansion models Sun et al. (2017), Shen et al. (2020). Several studies have contributed valuable insights into managementand market-level carbon optimization (Wang et al. 2022a, Sang et al. 2024, Lu et al. 2023). In contrast, this paper focuses on the practical implementation and evaluation of a carbon-aware operational strategies in a real-world industrial context. Specifically, we explore how on-site conventional power generation in a steel manufacturing plant can be scheduled to align with periods of lower carbon intensity on the regional electricity grid. Our case study focuses on the Bremen industrial cluster in Germany, an area characterized by a strong de-carbonization policy framework adopted by the Bremen Climate Protection and Energy Act (NACAO 2023). By dynamically adjusting the output of combined heat and power (CHP) units based on carbon intensity signals, we aim to demonstrate the carbon reduction potential of a emission-aware operational strategy. To achieve this, we designed a co-simulation framework integrating weather-dependent renewable generation, carbon intensity signals, real-world electricity consumption profiles, and operational characteristics of local generation assets. Through full-year, hourly resolution simulation experiments, we systematically evaluated the carbon footprint reduction achievable under different carbon-aware scheduling rules within different scenarios. This paper contributes to the growing literature on carbon-aware power systems by: •Demonstrating the technical feasibility of a rulebased carbon-aware scheduling strategy in a real industrial environment and highlighting the practical relevance of aligning power plant operations with low-carbon grid conditions using carbon intensity signals. •Validating the approach using data-driven cosimulation on an industrial grid model with multiple uncertainties (weather, load profiles, carbon 188 intensity, fuel type) and providing methodological guide for evaluating more complex carbon-aware strategies using co-simulation framework and datadriven scenarios. The findings suggest that even modest adjustments in local generation dispatch, guided by real-time carbon intensity data, can lead to measurable emissions reductions, without requiring radical infrastructure changes or complex control schemes. These results offer new operational possibilities for industrial de-carbonization and provide a foundation for future work integrating economic, market-based, and multi-energy considerations. The remainder of the paper is structured as follows: the next section outlines current trends in the carbon footprint associated with electricity generation and introduces the fundamentals of the carbon-aware power plant operation approach. The Method and Data section details the data sources, introduces the co-simulation approach, the high-level scenario definitions, and the evaluation methodology. The final sections present the results, followed by the conclusion and directions for future work. RESEARCH CONTEXT AND PRELIMINARY ANALYSIS Carbon-aware Approach for Power Systems In response to the urgent need to reduce the carbon footprint of power systems, a growing body of recent research has explored the integration of carbon emission signals, constraints, or models across various systemic levels of electricity generation and distribution — from strategic planning to real-time operation and control. For example, Sun et al. (2017) proposed a transmission planning method that quantifies consumptionbased carbon equity using an emission allocation index derived from carbon flow analysis. This approach clarifies consumers’ responsibility for system-wide carbon emissions. The method was tested on Garver’s 6-bus system and a modified IEEE 39-bus system, demonstrating improved equity performance. Another planning-level solution is presented in Shen et al. (2020), who developed a multi-objective power system transition model for scheduling the retirement of aging coalfired power plants. The model aims to minimize userside carbon footprints by balancing three conflicting objectives: cost, risk, and emissions. Validation was conducted using modified IEEE 24-bus and 118-bus systems. Several studies have also developed carbon-aware approaches for power flow optimization and equitable electricity distribution. For instance, Chen et al. (2024) introduced an optimal power flow model that incorporates carbon emission flow equations, constraints, and objectives, enabling co-optimization of electricity and carbon flows. This carbon-aware model extends conventional optimal power flow methods. Numerical simulations on a modified New England 39-bus system showed that the approach effectively coordinates diverse energy resources for grid de-carbonization. Similarly, Wang et al. (2024) proposed a novel method for calculating nodal carbon intensity, particularly suitable for systems incorporating electricity storage. Using a nondominated sorting genetic algorithm, they analyzed a 24-hour port simulation scenario with hourly resolution, integrating wind and solar power, electric storage, and hydrogen conversion technologies. Results confirmed the method’s potential for effectively reducing carbon emissions within port-related scenario. Carbon-flow-based methodologies have also been applied to power system scheduling, e.g. (Wang et al. 2022a), energy management frameworks (Sang et al. 2024, Wang et al. 2022b), and peer-to-peer carbon–electricity trading schemes (Lu et al. 2023). Of particular relevance is the work by Wang et al. (2022a), who developed a hybrid scheduling model incorporating data-driven carbon emission flows. Their model uses real-time carbon intensity monitoring to incentivize demand-side carbon reductions through equitable financial compensation. The approach was tested on IEEE 9-, 30-, 57-, and 118-bus systems. While these studies offer valuable insights into carbonaware strategies, they often rely on theoretical models or simplified IEEE test systems. In contrast, our study validates a carbon-aware operational strategy in a realistic industrial smart grid setting. Focusing on an industrial use case where carbon reduction is driven by local generation. We used a co-simulation framework to model the electrical infrastructure of the Bremen industrial cluster. This model incorporates locally distributed renewable generation, real-time carbon intensity estimation, and the operational profiles of on-site natural gas and metallurgical waste gas-fired power plants. A fullyear operational scenario was constructed to evaluate the technical feasibility of the proposed approach. Carbon Intensity and Energy Mix: Global Trend and Regional Potential Carbon intensity (CI) of electricity generation refers to the amount of carbon dioxide (CO2) emitted per kilowattor megawatt-hour (MWh) of electricity produced. It serves as a key indicator of the environmental impact of a country’s energy mix. While global average CI has been gradually decreasing due to worldwide sustainability efforts (see Fig. 1), the decline is not occurring fast enough to meet the targets necessary to limit global warming to 1.5°C (Ember 2024). Several countries have demonstrated a strong commitment to decarbonization through dedicated policies and long-term planning. For example, Germany’s Renewable Energy Sources Act (Erneuerbare-Energien-Gesetz) 189 2010 2012 2014 2016 2018 2020 2022 350 375 400 425 450 475 500 525 550 carbon intencity of electricity production, kg/MWh Germany World Figure 1: Carbon Intensity of Electricity Production in Germany and Worldwide (kgCO2/MWh, yearly average) Note. Sourced from ourworldindata.org (Ember 2024). sets ambitious goals, aiming for renewable energy to account for 65% of gross electricity consumption by 2030, with the ultimate objective of achieving a greenhouse gas-neutral electricity supply by 2050. Over recent decade, Germany’s CI has shown a significant downward trend. By 2023, the country had reduced its CI by 136 kgCO2/MWh, driven by the growing share of renewables and the corresponding decline in fossil fuel use (see Fig. 1). CI estimates are inherently tied to the energy mix which highly varies by region and over time. For instance, during sunny periods in areas with substantial photovoltaic (PV) adoption, solar energy becomes a larger share of the mix, resulting in lower CI. In contrast, regions reliant on fossil fuels experience higher CI. Furthermore, importing electricity from regions with low-CI generation can also significantly reduce local CI, often achieving lower emissions than local fossil-fuel-based generation. A clear example is the Niedersachsen region of Germany. Over the past three years, the region’s average annual CI has remained below the average value for conventional gas fuel combustion, even considering a conservative estimate of 270 kgCO2/MWh for natural gas-fired generation (Noussan et al. 2024), see Fig. 2. Fig. 2 shows that most of the data points representing daily average CI fall below the CI range for natural gas (indicated by the blue rectangle), and none fall within the CI range for blast furnace gas (indicated by the gray rectangle). This suggests that, for the majority of the time, the average CI of electricity supplied by the regional grid is lower, and thus more sustainable, than that of electricity generated using natural gas. Moreover, historical data show that more than 80% of the hours annually the regional grid had a CI below this threshold. Even with a stricter threshold, 50% lower Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 102 103 carbon intencity of electricity production, kg/MWh ( log scale) Carbon intensity (2021) Carbon intensity (2022) Carbon intensity (2023) Natural gas Blast furnace gas Figure 2: Carbon Intensity of Electricity Production in the Niedersachsen Region (kgCO2/MWh, log scale) Note. Sourced from co2map.de (Sundblad et al. 2023). Daily (dots), yearly (lines) averages and bound estimates (rectangles) for generation from natural gas — [270, 304] kgCO2/MWh and metallurgical waste BF/BOF gas — [794, 1102] kgCO2/MWh (Noussan et al. 2024, Messagie et al. 2013). than the natural gas lower bound, the regional grid still maintained the average CI below this level for over 50% of the hours per year, see Fig. 3). In this figure, the average proportion of hours per year is calculated over the period 2021–2023. On average, there are no hours during which the grid’s CI exceeds the upper threshold for metallurgical waste gases from Blast Furnace/Basic Oxygen Furnace (BF/BOF) processes, even when this threshold is reduced by 60%. These findings suggest that in such regions, a significant extra reduction in the total carbon footprint of the local power grid is theoretically achievable by implementing specific o perational s trategies. T hose s trategies imply reducing the runtime of local fossil-fuel power plants, especially during periods when low-CI electricity (either from surrounding renewables or imported) is available. Carbon-aware Power Plant Operation Strategy The core idea behind the discussed carbon-aware power plant operation strategy is to optimize the operation of local power plants in a way that minimizes total carbon emissions by aligning local electricity generation with periods of low carbon intensity on the regional grid. Rather than focusing on demand-side carbon reduction, this approach integrates carbon intensity monitoring into the operational decision-making of local generation units. Preliminary analysis above suggests that this strategy is both practical and feasible in real-world settings. Key principles of a carbon-aware approach towards power plant operations include: •Dynamic scheduling. Operating flexible conven190 10 20 30 40 50 60 70 80 90 100 bound-based threshold, % 10 20 30 40 50 60 70 80 90 100 percent of hours per year when the threshold is reached, % Blast furnace gas (upper bound) Blast furnace gas (lower bound) Natural gas (upper bound) Natural gas (lower bound) Figure 3: Hourly Budget of Low Carbon Intensity Electricity for the Niedersachsen Region Note. Calculated based the data sourced from co2map.de (Sundblad et al. 2023). Hourly polynomial approximation (lines) with 95% confidence intervals (shaded area) for the years 2021–2023, considering bound-based thresholds for natural gas — [270, 304] kgCO2/MWh and metallurgical waste BF/BOF gas — [794, 1102] kgCO2/MWh (Noussan et al. 2024, Messagie et al. 2013). tional power plants more during times when renewable energy supply is low (i.e., when marginal emissions are higher), and reducing output when clean energy is abundant. •Emission-aware dispatch. Prioritizing generation units with lower specific emissions over highemission ones, even if they are slightly more expensive to run. •Carbon signals integration. Utilizing real-time carbon intensity signals (e.g., from carbon tracking platforms) as inputs to plant control systems or scheduling software. To be able to evaluate such an approach we defined a simplified operation strategy as a set of scheduling rules for a conventional power plant, which are triggered when the estimated carbon intensity of the power grid falls below a specified threshold m. In such cases, the scheduled power output of the plant is reduced by a factor f, therefore leading to a reduction in the total carbon emissions associated with local electricity generation. The scheduling rules are defined as in Eq. 1: ft=        1−rif igrid t ifuel t ≤1−m ..., 1otherwise ,(1) where igrid tis the estimated carbon intensity of the power grid at time t, and ifuel tis the estimated carbon intensity of the power plant based on its fuel mix. In this example, the threshold mdefines the acceptable relative level of grid carbon intensity below which certain scheduling rule is triggered. Once triggered, the corresponding power output reduction factor ftat time tis determined by the rule value r. For example, if m= 0.1, r = 0.05 and the ratio igrid t/ifuel tis less than or equal to 0.9, then the plant’s scheduled power output at time twill be reduced by 5%. METHOD AND DATA Use-case: the Bremen Industrial Area As previously highlighted, the regions of Niedersachsen and Bremen offer unique conditions for testing the proposed approach. These areas already exhibit relatively low estimates of power grid carbon intensity, and the electricity supply for the Bremen industrial area is sourced from local conventional power plants operated by ArcelorMittal Bremen, GmbH. The industrial landscape of Bremen is currently undergoing a significant transformation, centered around the ArcelorMittal steel plant. This transformation is part of a broader initiative on decarbonizing heavy industry through the adoption of green hydrogen and renewable energy sources (ArcelorMittal 2024). A distinctive feature of this transition is the integration of local renewable energy generation, such as on-site photovoltaics and wind power, into a smart power grid designed to meet the substantial energy demands of steel production (ArcelorMittal 2024). The fully integrated steel plant is powered by an internal medium-voltage electricity network and conventional power generation (Fig. 4). Currently, the plant is a major source of carbon emissions, producing approximately six million tonnes of CO2annually, an amount comparable to the total emissions of the remaining part of Bremen (ArcelorMittal 2021). Consequently, the Bremen industrial area presents an ideal case study for exploring sustainable industrial practices and diverse power plant operation strategies aimed at reducing the carbon footprint of both steel manufacturing and energy generation. It may also serve as a model for similar de-carbonization efforts in heavy industries around the world. For this study, we utilized the validated hyBit open data collection (Petersen et al. 2025), which includes annual electricity consumption profiles tailored to the Bremen steel plant, historical weather data for Bremen, local photovoltaic (PV) system configurations, and specifications for onshore wind turbines in the surrounding industrial district (Fig. 4). The integrated highand medium-voltage electrical grid, consisting of 188 buses, is modeled using the pandapower library. The configurations for actual and potential PV systems and wind turbines around the steel plant 191 Figure 4: The Bremen Industrial Area Note. Schematic representation of the modeled capacity of photovoltaic (yellow icons), wind (blue icons), and conventional (purple icons) power plants around the Bremen steel plant (gray icons). are provided as comma-separated value (CSV) files in (Petersen et al. 2025). The steel plant electricity consumption is supplied by two conventional power plants: the Mittelsbüren natural gas combined heat and power (CHP) plant, with a net capacity of 445 MW, and a smaller blast furnace (BF) gas-fired power plant with a capacity of 246 MW (ArcelorMittal 2019). The carbon intensity of electricity production in the Bremen region was sourced using the API provided by the co2map.de project. This online platform offers hourly estimates of electricity-related CO2 emissions across german regions. It draws on data from sources such as ENTSO-E, Elexon BMRS, and EirGrid to analyze electricity generation, consumption, and cross-border flows (Sundblad et al. 2023). Evaluation Approach and Scenarios To evaluate the proposed strategy within the chosen use case, it is essential to design realistic scenarios. Given the heterogeneous set of models involved, including mediumand high-voltage power grid models, PV plants with varying configurations, wind turbines of different types, and diverse time-based data sources such as weather data and carbon intensity estimates, a cosimulation approach is well-suited for this task (Mihal et al. 2022). Co-simulation provides an effective means of orchestrating both time-based and event-driven simulation models (Ofenloch et al. 2022). In particular, we employed the mosaik framework, which has proven to be highly effective i n handling the complexity of real-world electricity grids integrated with distributed renewable energy sources and highresolution time series data. As modern energy systems become increasingly decentralized and data-driven, mosaik supports their modeling, analysis, and enables scenario based optimization approach (Ofenloch et al. 2022). The framework enables modular and scalable integration of heterogeneous simulation models through a flexible scenario API (Steinbrink et al. 2019), allowing for accurate representation of multi-domain knowledge, component interactions and temporal dynamics. Using the standard mosaik scenario definition, we developed two co-simulation scenarios, see also Fig. 5. Scenario A, the baseline, assumes that the industrial power grid (modeled using pandapower) is capable of meeting the electricity demand of the Bremen steel plant. This is achieved by prioritizing local renewable energy sources, followed by local conventional combined heat and power (CHP) plants with a total capacity of 700 MW (as described in the use case above). The local renewable sources are represented by 21 wind turbine models and 21 photovoltaic (PV) models (Fig. 4), with detailed configurations and power curves available in the referenced open repository, see Petersen et al. (2025). The electricity consumption of the steel plant is modeled here using two distinct demand profiles, each representing different weekly patterns based on operational modes determined by steel demand and production technology Petersen et al. (2025). Both profiles are scaled to match an annual energy consumption of 1.15 TWh (average annual value for the industrial area of Bremen), with hourly resolution. The baseline scenario further assumes that, in the absence of local renewable generation, the full electricity demand of the steel plant can be met by the local CHP units. These units can be fueled by a combination of natural gas and metallurgical waste gases, primarily blast furnace gas. Since there is no available information on the exact gas mixtures used in the CHP units, we performed calculations using fuel-bound estimates. Specifically, we considered a lower and upper bound for natural gas — 270 and 304 kgCO2/MWh, respectively (Noussan et al. 2024), and a lower bound for blast furnace gas at 794 kgCO2/MWh (Messagie et al. 2013). Scenario B, the experimental scenario, builds upon the same assumptions as the baseline. However, in this case, the CHP units are operated using the carbonaware scheduling strategy described earlier (see Eq. 1). To approximate the characteristic curve of this strategy, we defined a set of varying carbon intensity thresholds m∈ {1,3,5,7,10,13,15}. For simplicity, we also defined corresponding scheduling reduction rules r∈ {1,3,5,7,10,13,15}, such that m=rfor each experimental run. This means that when the ratio of grid to fuel carbon intensity igrid t/ifuel tis less than or equal to 1−m, the total power output of the CHP plants at time tis reduced by the same ratio 1−r. With this design, only seven com192 binations of thresholds and scheduling rules are tested to evaluate the feasibility of the proposed strategy. Both scenarios were simulated over a full year (360 days) with an hourly resolution. Since we have three weather reference years (2021-2023), which may differently affect local renewable generation, and two distinct electricity consumption profiles for the steel plant, we chose to simulate several parameterized combinations of the baseline and experimental scenarios. Also, the average relative difference between scenario results must be calculated separately for each carbon intensity bound. This resulted in 42 parametrized scenario runs per bound or 252 full year simulations overall (see Table 1). Table 1: Scenario Configuration Parameters Scenario A Scenario B r,m,%- 1, 3, 5, 7, 10, 13, 15 CHP units 2 PV models 21 WT models 21 Reference weather years 2021, 2022, 2023 Reference steel plant load profiles A, B Carbon intensity bounds, kgCO2/MWh 270, 304, 794 Note. Carbon intensity bounds for natural gas and metallurgical waste gas are defined based on (Noussan et al. 2024, Messagie et al. 2013). Our evaluation approach is based on calculating the total indirect carbon emissions associated with local electricity production and consumption. These emissions are computed as the sum of those resulting from power drawn from the grid and from on-site generation, see Eq. 2. Etotal =X t Plocal t×ifuel t+Pgrid t×igrid t,(2) here Plocal tand Pgrid trepresent the electricity supplied by local power plants and the external grid at time t, while ifuel tand igrid tdenote their respective carbon intensity estimates. It is clear that when the grid is supplied by green energy, the corresponding igrid tis considered to be zero, and the total carbon footprint over time is determined solely by fuel-based emissions. RESULTS Following the evaluation approach outlined in the Method section, we executed 252 parameterized coWeather data PV Simulator Carbon intencity estimates Output data WT Simulator Steel Plant profiles Power Plant Simulator Power Grid Simulator Control Simulator Figure 5: The Co-simulation Scenario Setup Note. It illustrates the data flow (solid) and control signals (dashed) between models and data entities. PV stands for photovoltaics and WT stands for wind turbine simulation models. simulations across different reference weather years, the steel plant profiles, a nd e mission f actor r eference estimates (see Table 1). For the purpose of scenario comparison, the relative difference between the total carbon footprints (see Eq. 2) of each parametrized pair of AB scenarios must be calculated, allowing the separate analysis of each fuel carbon intensity bound. Thus, this computational effort r esulted i n 1 26 difference points of the total carbon footprints (Etotal) for corresponding scenarios. These differences ( in percentage points) were then grouped by related fuel carbon intensity bounds (upper/lower bounds for natural gas and blast furnace gas), averaged and plotted against the corresponding scheduling rules/thresholds (also in percentage points), see Fig. 6. Exact numerical results are shown in Table 2. As shown in Fig. 6, even a simplified d efinition of a carbon-aware operational strategy proves effective when applied to historical data and the designed scenarios. For example, when the scheduling rule is triggered at the point where the grid’s carbon intensity estimate falls below the lower bound for natural gas (gray line, Fig. 6), a 10% threshold results in an average carbon footprint reduction of approximately 4.5%. In contrast, when the CHP units are assumed to use a gas mixture with a high share of blast furnace gas — i.e., the lower bound for blast furnace gas is used as the threshold base (brown line, Fig. 6) — the average difference of the same 10% threshold is significantly higher, reaching around 8%. In our scenario setup, a 10% threshold leads to an annual carbon footprint re193 Table 2: The Scenarios Evaluation Results Carbon intensity bound Results Scheduling rule / threshold 1% 3% 5% 7% 10% 13% 15% Blast furnace gas lower bound ∆, % 0.80 2.41 4.01 5.62 8.02 10.43 12.03 ∆, ktCO26.98 20.94 34.90 48.86 69.80 90.74 104.69 Natural gas upper bound ∆, % 0.50 1.51 2.51 3.51 5.00 6.47 7.44 ∆, ktCO21.68 5.03 8.39 11.72 16.70 21.63 24.87 Natural gas lower bound ∆, % 0.46 1.36 2.27 3.17 4.52 5.85 6.72 ∆, ktCO21.35 4.05 6.74 9.43 13.42 17.36 19.96 Note. Absolute (∆, ktCO2) and relative (∆, %) differences in total carbon footprints across scenarios are averaged over reference weather years and electricity consumption profiles. 1 3 5 7 9 11 13 15 scheduling rule / threshold, % 1 3 5 7 9 11 13 15 difference in total carbon footprint, % Blast furnace gas (lower bound) Natural gas (upper bound) Natural gas (lower bound) Figure 6: The Scenarios Evaluation Results Note. Relative differences in total carbon footprints across scenarios (dots) are averaged over reference weather years and electricity consumption profiles using linear regression (colored lines). The dashed line represents ideal case when zero-emission electricity supplied by the grid. Exact numerical results are shown in Table 2. duction of approximately 5-8%, which corresponds to roughly 13.4-69.8 ktCO2 per year. The dashed diagonal line in Fig. 6 represents an idealized, though sofar unrealistic, case in which the grid supplies entirely zero-emission energy. In such a case, the threshold value would be equivalent to the reduction value according to Eq. 2. The results also suggest an approximately linear relationship between the scheduling rule / threshold and the resulting relative carbon footprint reduction within the use case. The slope of this relationship depends on both the grid’s carbon reduction potential, local CHP units flexibility, a nd t he availability o f l ocal renewable energy. It is important to note, however, that more complex combinations of rules, where m = r, may result in a more intricate solution space when determining the strategy’s characteristic curve. This added complexity could significantly increase the computational cost of simulation and optimization, and is therefore left for future research. DISCUSSION This study contributes to the growing body of carbonaware energy management research by validating a simplified operational strategy within a real-world industrial context, contrasting with many prior works that rely on theoretical models or simplified test systems. For example, while Sun et al. (2017) and Shen et al. (2020) developed carbon-aware planning models using the IEEE 24-/118-bus systems, these models remain largely abstract and far from industrial practice. Similarly, recent studies such as Chen et al. (2024) and Wang et al. (2024) have proposed technically advanced carbon-aware models including carbon/power flow optimization and nodal carbon intensity estimates, but tested them primarily in synthetic or small-scale system environments. In contrast, our study targets a practical and industrially relevant setting: the Bremen steel plant, a highemission, large-scale industrial cluster. By utilizing a co-simulation framework, we demonstrate the technical feasibility and measurable carbon reduction potential of carbon-aware operation strategies under realistic constraints. This distinguishes our work from earlier research (e.g., Wang et al. (2022a)) which emphasized theoretical scheduling benefits but lacked industrial validation and often relied on idealized or real-time carbon monitoring without coupling to actual operational models. Moreover, our approach offers a pragmatic balance between simplicity and effectiveness. Instead of optimizing complex multi-objective formulations or relying on computationally intensive algorithms (such as genetic algorithms used in Wang et al. (2024)), we implement a threshold-based strategy that scales well and remains interpretable for industrial operators. Importantly, our results indicate that even simplified carbon-aware operational rules, such as reducing local fossil-fuel generation by 10% during periods of green energy injections and lower grid electricity, can yield significant annual emissions reductions (up to 8% de194 pending on the fuel mix). These gains are particularly notable in regions like Niedersachsen and Bremen, where the carbon intensity of grid electricity often undercuts the emission factors of on-site fossil generation. Thus, the study demonstrates how regional energy characteristics can be leveraged to align industrial operations with broader de-carbonization targets. While the results are promising, their practical implementation depends on external factors including grid operator policies, electricity market design, and demandside flexibility. This study focuses on technical feasibility and does not address economic or regulatory dimensions in detail. Finally, the use of a co-simulation environment enables future exploration of more sophisticated strategies, such as asymmetric scheduling rules, hybrid storage integration (e.g., batteries, hydrogen), and market-coupled dispatch models. These directions, while computationally and methodologically more demanding, are essential for developing more resilient and economically viable carbon-aware solutions. CONCLUSION AND FUTURE WORK The results prove that carbon-aware operational strategies are feasible within the context of a flexible industrial smart grid integrating carbon intensity monitoring and the overall downward trend in electricity carbon intensity observed in the regional power system. A scheduling rule in the range of 10–15% can also be considered technically feasible from both grid capacity and grid operator perspectives. However, its practical implementation remains subject to demand-side management strategies and broader economic considerations. The study doesn’t just propose an exact operating rules for a carbon-aware strategy but shows technical feasibility and measurable carbon benefits under different realistic configurations, which many prior studies lack. Unlike prior studies, this work validates the proposed methodology on the real-world industrial use-case - Bremen steel plant, which is a major emitter and represents a typical large-scale industrial setup. This study may also serve as a methodological guide for evaluating more complex carbon-aware strategies and models using a co-simulation framework and data-driven scenarios. Since the primary aim of this paper was a technical evaluation, we intentionally leave economic and market considerations for future research. Thus, the proposed operational strategy demonstrated to be able to support broader climate goals by lowering the carbon footprint of electricity production, facilitating greater integration of renewable energy, and aligning power plant operations with de-carbonization pathways. It is especially relevant for industrial facilities with onsite generation and grid-connected flexibility, such as CHP plants and energy hubs. However, future work should explore more sophisticated, context-specific strategies, incorporating market signals and greater operational flexibility, including the integration of industrial batteries and hydrogen-topower units. ACKNOWLEDGEMENTS This research is a part of hyBit project, funded by German Federal Ministry of Education and Research (FKZ 03SF0687D). Data and Code Availability All data, developed components and algorithms are referenced or can be found in the public repository: https://gitlab.com/hybit1/hybit-mix-matters REFERENCES ArcelorMittal, 2019. Power-GIS mit Wachstumspotenzial. ArcelorMittal Bremen GmbH. URL https://www.mu m.de/-/media/mum/downloads/referenzen/mum-kun denreferenz-arcelor-mittal-092019.pdf [Accessed: 15.05.2025]. ArcelorMittal, 2021. ArcelorMittal plans major investment in German sites, to accelerate CO2 emissions reduction strategy and leverage the hydrogen grid. ArcelorMittal. URL https://corporate.arcelormittal.com/media/ne ws-articles/arcelormittal-plans-major-investmen t-in-german-sites-to-accelerate-co2-emissions-r eduction-strategy-and-leverage-the-hydrogen-grid [Accessed: 15.05.2025]. ArcelorMittal, 2024. Transformation in Bremen and Eisenhüttenstadt. ArcelorMittal Europe Communications. URL https://europe.arcelormittal.com/newsandmedia/pr essreleases/press-2024/transformation-Bremen-Eis enhuttenstadt [Accessed: 15.05.2025]. Chen X., 2024. Enhance low-carbon power system operation via carbon-aware demand response.Energy Internet, 1, no. 2, 141–149. doi:10.1049/ein2.12004. Chen X.; Sun A.; Shi W.; and Li N., 2024. Carbon-Aware Optimal Power Flow.IEEE Transactions on Power Systems, 1–14. doi:10.1109/TPWRS.2024.3514516. Dagnachew A.; Hof A.; van Soest H.; and van Vuuren D., 2021. Climate change measures and sustainable development goals. PBL Netherlands Environmental Assessment Agency. URL https://www.pbl.nl/en/publications/c limate-change-measures-and-sustainable-developme nt-goals [Accessed: 15.05.2025]. Ember, 2024. Carbon intensity of electricity generation – Ember and Energy Institute. OurWorldinData.org [dataset]. URL https://ourworldindata.org/grapher/c arbon-intensity-electricity [Accessed: 15.05.2025]. IEA, 2024. Real-time data can help to track and reduce emissions from electricity sector. IEA, Paris. URL https: 195 //www.iea.org/commentaries/real-time-data-can-h elp-to-track-and-reduce-emissions-from-electrici ty-sector [Accessed: 15.05.2025]. Lu Z.; Bai L.; Wang J.; Wei J.; Xiao Y.; and Chen Y., 2023. Peer-to-Peer Joint Electricity and Carbon Trading Based on Carbon-Aware Distribution Locational Marginal Pricing.IEEE Transactions on Power Systems, 38, no. 1, 835–852. doi:10.1109/TPWRS.2022.3167780. Messagie M.; Boureima F.; Mertens J.; Sanfelix J.; Macharis C.; and Mierlo J.V., 2013. The Influence of Allocation on the Carbon Footprint of Electricity Production from Waste Gas, a Case Study for Blast Furnace Gas.Energies, 6, no. 3, 1217–1232. doi:10.3390/en6031217. Mihal P.; Schvarcbacher M.; Rossi B.; and Pitner T., 2022. Smart grids co-simulations: Survey & research directions. Sustainable Computing: Informatics and Systems, 35, 100726. ISSN 2210-5379. doi:10.1016/j.suscom.2022. 100726. NACAO, 2023. Catalogue of Good Practices & Mapping of Carbon Sinks. Free Hanseatic City of Bremen. NACAO, Interreg Europe. URL https://www.interregeurope.e u/sites/default/files/2024-05/NACAO_Situationa l%20Study%20Bremen%20%28Germany%29.pdf [Accessed: 15.05.2025]. Noussan M.; Fambri G.; Negro V.; and Chiaramonti D., 2024. Hourly electricity CO2 intensity profiles based on the real operation of large-scale natural gas combined cycle cogeneration plants.Energy, 312, 133424. ISSN 0360-5442. doi:10.1016/j.energy.2024.133424. Ofenloch A.; Schwarz J.S.; Tolk D.; Brandt T.; Eilers R.; Ramirez R.; Raub T.; and Lehnhoff S., 2022. MOSAIK 3.0: Combining Time-Stepped and Discrete Event Simulation. In Open Source Modelling and Simulation of Energy Systems (OSMSES). IEEE, 1–5. doi:10.1109/OSMSES54 027.2022.9769116. Petersen K.; Valko D.; and Tolk D., 2025. hyBit open data collection. Zenodo. doi:10.5281/zenodo.15052032. URL https://gitlab.com/hybit1/hybit-repo. Sang L.; Xu Y.; and Sun H., 2024. Encoding Carbon Emission Flow in Energy Management: A Compact Constraint Learning Approach.IEEE Transactions on Sustainable Energy, 15, no. 1, 123–135. doi:10.1109/TSTE.2023.3274 735. Shen W.; Qiu J.; Meng K.; Chen X.; and Dong Z.Y., 2020. Low-Carbon Electricity Network Transition Considering Retirement of Aging Coal Generators.IEEE Transactions on Power Systems, 35, no. 6, 4193–4205. doi: 10.1109/TPWRS.2020.2995753. Steinbrink C.; Blank-Babazadeh M.; El-Ama A.; Holly S.; Lüers B.; Nebel-Wenner M.; Ramírez Acosta R.P.; Raub T.; Schwarz J.S.; Stark S.; Nieße A.; and Lehnhoff S., 2019. CPES Testing with mosaik: Co-Simulation Planning, Execution and Analysis.Applied Sciences, 9, no. 5. ISSN 2076-3417. doi:10.3390/app9050923. Sun Y.; Kang C.; Xia Q.; Chen Q.; Zhang N.; and Cheng Y., 2017. Analysis of transmission expansion planning considering consumption-based carbon emission accounting.Applied Energy, 193, 232–242. ISSN 0306-2619. doi: 10.1016/j.apenergy.2017.02.035. Sundblad M.; Fürmann T.; Weidlich A.; and Schäfer M., 2023. Load and generation time series for German federal states: Static vs. dynamic regionalization factors. In Open Source Modelling and Simulation of Energy Systems (OSMSES). IEEE, 1–6. doi:10.1109/osmses58477.2023.1 0089686. Wang W.; Huo Q.; Liu Q.; Ni J.; Zhu J.; and Wei T., 2024. Energy Optimal Dispatching of Ports Multi-Energy Integrated System Considering Optimal Carbon Flow.Trans Intell Transport Sys, 25, no. 5, 4181–4191. ISSN 15249050. doi:10.1109/TITS.2023.3336998. Wang Y.; Qiu J.; and Tao Y., 2022a. Optimal Power Scheduling Using Data-Driven Carbon Emission Flow Modelling for Carbon Intensity Control.IEEE Transactions on Power Systems, 37, no. 4, 2894–2905. doi:10.1109/TP WRS.2021.3126701. Wang Y.; Qiu J.; and Tao Y., 2022b. Robust energy systems scheduling considering uncertainties and demand side emission impacts.Energy, 239, 122317. ISSN 03605442. doi:10.1016/j.energy.2021.122317. 196