How we modelled 24/7 CFE and green hydrogen for India’s steel sector

Summary

We extended TransitionZero's PyPSA model of India to test 24/7 carbon-free electricity (CFE) and green hydrogen blending for the steel sector, using electricity and hydrogen demand profiles derived from the projected 2030 capacity of electric arc furnaces (EAF), induction furnaces (IF), and direct-reduced iron (DRI).

A scenario comparison approach lets us isolate the different cost drivers of CFE and green hydrogen, including the difference between standalone and co-optimised scenarios.

System-cost optimality is not the same as plant-level optimality. We layer state-specific tariffs and subsidies on top of our optimisation model to estimate the real delivered costs of electricity, hydrogen, and steel.

Introduction

Our companion blog showed that 24/7 CFE and green hydrogen can decarbonise much of India's electric furnaces and natural gas DRI shafts for less than a 10% uplift on the cost of steel. Running both of these policies together can lead to co-optimisation benefits, including lower system costs and reduced renewables curtailment.

This blog provides an overview of the methodology. It covers the model of India we used, how we estimated the steel sector's electricity and hydrogen demand, how green hydrogen was represented in standalone and co-optimised forms, and how we translated system-level results into a cost uplift a steelmaker would actually face. The full model and methodology is released under MIT licence, available to download here.

Our model of India

We use PyPSA, an open-source dispatch and capacity-expansion framework, to represent the whole of India as five interconnected transmission grid zones: East, North, North-East, South, and West. Each zone has its own demand profile, generation fleet, and renewable resource. Interconnector capacity between zones is fixed based on the Central Transmission Utility's 2028-29 plan. No intra-zone flows are modelled.

The model optimises generation and dispatch at hourly resolution for a single year, 2030. Our starting point is a Nationally Determined Contributions (NDC)-compliant 2030 grid: 50% non-fossil installed capacity, 500 GW of non-fossil capacity, and a 45% reduction in GDP emissions intensity against 2005. The remaining generation capacity is built at least cost to meet demand according to the above constraints, and taking into account known pipelines and build limits.

The full set of input assumptions, including capital costs, capacity factors, fuel prices, and build constraints, is published in the annex of the main report. Sources include the CEA-DEA Indian technology catalogue, IEA data, and TransitionZero's own analysis. We use 2022 prices in our modelling, along with 2022 exchange rates between USD and INR.

Modelling 24/7 CFE

The CFE framework follows the same methodology as our previous CFE work, where more information can be found on detailed modelling decisions. The code, data, and results, are publicly available on GitHub.

In adopting a CFE policy, a commercial and industrial (C&I) consumer (in this case the total EAF/IF and DRI demand in a given grid zone) must hit a target CFE score. The score is calculated each hour as the share of demand met by either contracted clean generation or by the carbon-free portion of the grid. The score is averaged across all 8,760 hours of the year.

Two changes distinguish this study from our earlier work. First, we use demand profiles specific to the steel sector in 2030 and which include hydrogen demand, rather than national load curves. Second, we focus on the three grid zones where the relevant capacity sits: West, South, and East. Demand profiles differ significantly between zones, reflecting the geographic distribution of the steel industry.

To meet the CFE target, the model optimises the buildout and dispatch of additional solar, onshore wind, and battery storage, contracted to the steelmaker via power purchase agreements (PPAs). This buildout must ensure that the steelmaker hits the target CFE score at lowest cost throughout the year, using a combination of CFE supply and grid supply. We allow some surplus generation to be sold back to the grid in times of generation excess.

Modelling the steel sector and deriving the load curves

The hourly electricity and hydrogen demand profiles are built from the bottom up for each grid zone. This process has six steps:

1. Capacity by region. We estimated 2030 EAF and IF capacity in each grid zone using Global Energy Monitor data and Ministry of Steel projections.
2. Production by route. We adopt the Ministry of Steel's 2030 production figures, split across EAF, IF, and DRI.
3. Regional and monthly allocation. Production is distributed to grid zones by their share of capacity, then split into monthly production using the 2024 monthly shape from the World Steel Association.
4. DRI capacity. Total DRI capacity, including the split between NG-DRI and coal-DRI, is calculated using Global Energy Monitor data and Ministry of Steel projections.
5. Hydrogen substitution. A chosen percentage (0%, 10%, 20%, or 40%) of NG-DRI is replaced with H₂-DRI. Coal-based DRI and scrap-fed routes have no associated hydrogen demand and are left unchanged.
6. Load curves. Electricity-use factors from the Ministry of Steel are applied to EAF, IF, and DRI production to build hourly electricity load curves. Hydrogen demand uses a factor of 58 kg H₂ per tonne of DRI.

This produces, for each grid zone, two outputs: an hourly electricity load curve for furnaces and DRI, and an hourly hydrogen demand figure (in millions of tonnes per annum) for the substituted share of NG-DRI.

How we modelled green hydrogen blending

We now have an hourly demand profile, per grid zone, for each blending ratio in our model (note that this is an additional load to the model, as it substitutes natural gas demand that is not included). We allowed the model to build the following technologies to meet this demand:

• Additional solar, onshore wind, and battery storage
• Electrolyser (Proton Exchange Membrane) and balance of plant
• Hydrogen storage (above ground steel tanks)

The optimisation can then choose whether to oversize the electrolyser to take advantage of surplus energy, and how much to invest in either battery storage or hydrogen storage (i.e. whether to store energy in electricity or hydrogen, considering electrolyser efficiencies and operating costs).

In a ‘standalone’ production scenario, the model simply builds to meet the hydrogen demand, and there is no interaction with CFE demand. This isolates the pure cost of meeting the hydrogen demand at lowest cost, assuming a group-captive setup.

Under a ‘co-optimised’ scenario, the model optimises CFE and hydrogen procurement jointly at lowest-cost. This allows resources to be shared across both uses, resulting in less buildout, and lower costs - the source of the co-optimisation benefit.

Comparing the standalone scenario to the co-optimised scenario allows us to discover the co-optimisation benefit, i.e. how much less needs to be built for green hydrogen production, because the CFE infrastructure can already be used.

How we estimated the final cost uplift to the levelised cost of steel

Our model returns a system-optimal solution, minimising the real economic resources consumed by the power sector as a whole. This does not translate directly into what a steelmaker might pay. Grid tariffs and additional charges, including inter-state transmission system (ISTS) charges, transmission and wheeling fees, are transfers between sectors of the system and net out on the system level. However, they must be accounted for when considering the landed cost of electricity and steel for steelmakers.

To estimate plant-level impacts, we layer on these estimated charges per grid zone, taking one representative state per grid zone: Gujarat in the West, Karnataka in the South, and Odisha in the East. We apply each state’s tariffs and subsidy regime for any renewable or green hydrogen projects. This includes the grid tariff for shortfall hours (a blend of captive, secondary PPAs, distribution company supply, and other sources), the export tariff for surplus, and any relevant waivers, subsidies, and charges.

The import tariffs represent blended residual electricity costs for steelmakers to cover CFE/PPA shortfall hours. They are intentionally lower than distribution companies’ (DISCOM) retail tariffs to reflect blending of different sources to reach a lowest-cost procurement strategy. These are regional averages, and plant-specific procurement may vary significantly. These would also vary by time of day, depending on the exact hours of PPA shortfall.

The change in the levelised cost of steel is then calculated by summing the incremental costs of the following on a regional level:

1. Hourly matching costs of the plant - the costs of building additional capacity (solar, onshore wind, battery) to meet CFE targets for all the EAF/IFs in the grid region, excluding the electrolyser.
2. Avoided electricity costs - avoided grid charges or other electricity costs, as a result of moving to a CFE supply.
3. Electrolyser capital costs - the cost of building and operating the electrolysers, balance of plant, and any hydrogen storage tanks.
4. Hourly matching costs for electrolyser - the costs of building additional capacity (solar, onshore wind, battery) to match electrolyser demand with clean energy on a 100% hourly basis.
5. Reduced natural gas consumption - resulting in avoided natural gas costs due to replacement by hydrogen in DRI.
6. Co-optimisation benefits - saved costs from allowing CFE supply and green hydrogen production to share procured renewable assets, reduce curtailment, and save on costs.

Each of these has an incremental cost or benefit, which can be divided by the total steel production in the region to give an estimate to the levelised cost of steel, in USD per tonne crude steel.

There are limitations to this approach. In reality, the tariff and subsidy impacts would feed back into the steelmaker procurement decisions, which would change the optimal capacity expansion - this plant-led decision making may differ from a system optimal solution.

Key assumptions and sensitivities

As with any modelling and forecasting study, there are key sensitivities to bear in mind that could change the outcomes of this study.

1. Existing grid tariffs and supply agreements - steel plants with a majority captive power supply will find it much costlier to switch to CFE, compared to a plant relying more on secondary PPAs and grid supplies. Whilst we have tried to capture regional averages here, the plant-by-plant situation would look very different.
2. The natural gas price - current geopolitical uncertainties have exposed the vulnerability of fossil fuel supply chains, and a world facing large supply shocks will see more benefit to switching to domestic green hydrogen supplies.
3. The ratio of NG-DRI to EAF/IF capacity in each grid zone - the ratio between furnace demand and hydrogen (and by proxy, electrolyser) demand shapes procurement strategies, total steel cost uplifts, and the value of co-optimisation. This can be seen in the varying results between grid regions.
4. The level of co-operation to facilitate sharing PPAs and renewable generation - our co-optimised scenarios assume no barriers between sharing procured electricity for hydrogen production and furnace load. Although this is straightforward for an integrated NG-DRI-EAF plant, this gets more complicated when thinking about standalone DRI or standalone furnaces.

What this means

The numbers in our blog and main report are the output of a nationwide system cost model, with a regulatory overlay. They reflect a possible future for the steel sector which chases ambitious decarbonisation targets, and the resulting changes to dispatch and capacity expansion decisions. It gives policymakers and those thinking about systemic changes an idea of the scale of the challenge, and the opportunities ahead.

We encourage you to look into the main report, as well as the methodology document of our original CFE study, to understand more of the detail around our modelling and analytical framework.

The model itself, along with input data and documentation, is released under MIT licence, available to download here.

This blog is the second in a two-part series based on research by Verity Crane, Heavy Industry Lead; Irfan Mohamed, Senior Energy Systems Modeller; George Ebri, Research Analyst; and Abhishek Shivakumar, Head of Solutions Engineering. Read part one here.

Green steel, hour by hour

Download the reportDownload the report