Natural gas has been the backbone of heating processes in small and medium-sized manufacturing for years. It’s reliable, familiar, and relatively cheap , but it’s also a big source of carbon emissions. Green hydrogen, made by splitting water with renewable electricity, sounds like an ideal replacement. No carbon, just water and energy. But that raises a practical question: could a small manufacturer actually make their own hydrogen on-site, using solar panels and batteries?

That’s what we studied in our, “Techno-economic pathway for green hydrogen adoption in thermal applications across US small and medium manufacturing sectors”. We modeled a system that includes solar PV panels to generate electricity, a battery to store energy when the sun isn’t shining, and an electrolyser to produce hydrogen. The goal was to figure out if it’s financially and technically feasible for small manufacturers to replace some or all of their natural gas with this on-site green hydrogen.

The model tried to find the “sweet spot” , the optimal sizes for the solar, battery, and electrolyser setup that would meet a factory’s hydrogen needs over 20 years while keeping costs as low as possible. We built this using Mixed-Integer Linear Programming (MILP), which let us juggle all the technical variables and financial assumptions to land on a design that balances performance and affordability.

Along the way, we had to represent each part of the system with math. For solar energy, we calculated hourly power production using this equation:

$$ P_{pv}(t) = f_{pv} \cdot P_{pv,rated} \cdot \left( \frac{G(t)}{G_{std}} \right) \cdot \left(1 + \alpha \cdot (T_{cell}(t) - T_{cell,std})\right) $$

This formula considers sunlight, temperature effects, and system losses to estimate how much electricity the panels produce each hour.

For the battery, we tracked how energy is stored and used over time with:

$$ E_{b}(t) = E_{b}(t-1) \cdot (1 - \sigma_{b}) + \eta_{b,ch} \cdot P_{b,ch}(t) - \frac{P_{b,dch}(t)}{\eta_{b,dch}} $$

It accounts for charging efficiency, losses during discharge, and even small self-discharge over time. We also optimized for battery size based on how much energy storage the system would need across seasons.

Hydrogen production was modeled using:

$$ \dot{m_{H_2}} = \frac{P_{el}(t) \cdot \eta_{el}}{HHV_{H_2}} $$

This let us estimate how much hydrogen the electrolyser can produce each hour based on the input power and its efficiency.

Then came the cost side. Our objective was to minimize the Net Present Cost (NPC) over 20 years, combining initial capital expenses (CAPEX), ongoing operational costs (OPEX), and the additional cost to retrofit burners so they could handle hydrogen. The overall cost function looked like this:

$$ NPC_{H2} = Capex_{PV} + Capex_{BESS} + Capex_{EL} + \sum_{n=1}^{N} \frac{Opex_{total,year\ n}}{(1 + r)^n} + Capex_{burner_conversion} $$

We layered on constraints to make sure the system made sense , the PV had to power the electrolyser and charge the battery, the hydrogen output had to meet the annual target, the battery couldn’t overcharge or fully drain, and everything had to stay within realistic size limits.

What the Results Actually Showed

What stood out right away was that larger hydrogen replacement targets required bigger solar arrays and batteries. That was expected. But interestingly, the size of the electrolyser didn’t scale much. Instead of going for a massive unit to cover peak demand, the system preferred a modestly sized electrolyser running more hours per year, powered by a bigger battery and PV setup.

The bad news? Cost. Replacing even 50% of a factory’s natural gas use with on-site hydrogen gets expensive quickly. When we calculated the Levelized Cost of Hydrogen (LCOH), it landed between $13.40 and $14.00 per kilogram , which is way higher than the equivalent cost of natural gas (closer to $1–$2/kg depending on prices). Most of this cost came from the solar panels, followed by the battery system.

We also tested how much of a carbon tax you’d need to make hydrogen financially competitive with natural gas. The numbers were eye-watering. To break even at 50% hydrogen replacement, you’d need a carbon price of around $700 per tonne of CO₂. For full replacement, that figure jumped to more than $2,000 per tonne , far beyond what’s likely in the near future.

Still, the results weren’t all discouraging. We found that starting small , say with 20% hydrogen replacement , could be a more practical first step. The systems are cheaper, easier to manage, and still help lower emissions. And if manufacturers can tap into financial incentives like grants or tax credits for solar and burner conversions, the economics start to look a lot better.

Another interesting finding was around electricity sourcing. Using grid electricity , especially if the grid is already renewables-heavy , might be a smarter move than building an entire solar-battery infrastructure. If the grid gets cleaner and electricity prices remain stable, that could dramatically cut costs.

In the end, on-site green hydrogen production for small manufacturers isn’t fantasy. It’s technically feasible. But unless the costs drop significantly or policy support ramps up, it’s still out of reach for most. Our model provides a clear picture of what it would take , financially, technically, and operationally , to make this work. Hopefully, that helps guide the next steps, whether that’s more research, new incentives, or just smarter ways of phasing in hydrogen, one kilowatt-hour at a time.

If you Want the full technical paper, You can read it here: https://doi.org/10.1016/j.ijhydene.2024.12.069. It’s behind the paywall though, didn’t had enough funding to open access it.