The Pros and Cons of Hydrogen Fuel Cells as Backup Generators

Microsoft has been experimenting with hydrogen power since 2013 when it began testing solid oxide fuel cells (SOFCs), which extract hydrogen (H2) from natural gas, to run server racks in data centers. SOFC technology was (and still is) expensive, relies on fossil fuels, and produces CO2, so the software giant eventually put the project on the back burner.

Meanwhile, the price of proton exchange membrane (PEM) fuel cells has dramatically decreased. PEM fuel cells use purified H2 rather than hydrogen extracted from natural gas. They operate at lower temperatures and are carbon neutral, making them favorable alternatives to SOFCs.

Microsoft's 250kW PEM fuel cell backup generator built by Power Innovations. (Image courtesy of Power Innovations.)

Instead of trying to run the data centers exclusively from fuel cell energy, Microsoft is using PEM fuel cells to replace diesel-powered backup generators, which are called into service so infrequently that they burn more fuel during monthly testing than they do in actual operation. But they still require maintenance and diesel fuel has a limited shelf life, so finding a reliable, cost-effective replacement makes sense. For a comparable price, PEM fuel cells can deliver the same generating capacity as diesel generators.

Fuel Costs

While the price of a PEM fuel cell is similar to that of an equivalent diesel generator, the cost of the fuel the systems use is not. One kilogram of hydrogen has almost as much energy as a gallon of diesel fuel, but per unit of energy content, hydrogen fuel costs more than five times as much as diesel. Some of that cost is offset by the fact that fuel cells are twice as efficient as diesel generators, but in terms of price per kilowatt hour of electricity generated, hydrogen is still nearly three times more expensive. If only there were an abundant, inexpensive source of hydrogen available. Oh, yeah … there is!

Electrolysis

Two-thirds of our planet’s surface is covered in a hydrogen-rich substance (H2O), but those hydrogen atoms are quite attached to their oxygen siblings, and covalent bonds are tough to break. Separating water molecules by electrolysis requires about 50kWh of electricity to obtain a single kilogram of hydrogen. (The resulting hydrogen contains 33kWh of energy.) That’s a problem if you’re buying electricity from the grid, but less of a concern if you’re using solar, like Microsoft plans to do. Here’s how electrolysis by solar works:

(Video courtesy of the U.S. Department of Energy.)

Fueling Up

Although the photovoltaic effect and the electrolysis process are both relatively inefficient, for this particular application—backup generation—efficiency is less important, for two reasons. First, sunlight is free. (Solar panels cost money, but that’s a small initial investment in this case.) Second, these generators rarely run, so their hydrogen tanks can be “trickle charged,” so to speak, with a relatively small array that runs the electrolyzer whenever the sun happens to be shining.

Hydrogen tanks. (Image courtesy of Microsoft.)

For example, Microsoft’s fuel cell, with a 250kW capacity, was able to power 10 racks of data servers for 48 hours. That’s a total of 12MWh of electricity. Fuel cells are about 60 percent efficient, so it would take 600kg of hydrogen to produce that much electricity.

How big of an array would you need to replenish that? Well, at 50kWh/kg, it takes 30MWh of electricity to produce 600kg of hydrogen. But keep in mind that these are backup generators; it’s extremely unlikely that they’ll be running for 48 hours straight. If you start with enough hydrogen to power the racks for a few hours, say, 60kg or so, then a one-megawatt solar array could produce the remaining 540kg in less than a week. According to Microsoft engineer Mark Monroe, the generators sit idle for more than 99 percent of their lives, which means they run about three days a year. On average, then, electrolyzers could take up to four months to slowly refill the hydrogen tanks. At an average of five peak sun hours per day, a modest 50kW array could do the job.

Why Not Use Batteries?

Looking at my calculations above, we see that it takes 50kWh of energy to extract one kilogram of hydrogen, which, in turn, can generate 20kWh of electricity. That’s a round-trip efficiency of 40 percent—pretty bad compared to batteries, which easily exceed 85 percent.

Before the days of cloud-based data centers, companies had on-premises servers with backup power provided by uninterruptible power supplies (UPSs) consisting of a charge controller, a battery, and an inverter. Many data centers continue to run with UPSs on their servers. If batteries work at the server level, then why not use them at the data center level?

When choosing an energy storage system, engineers have to consider factors such as initial price, operation and maintenance costs, round-trip efficiency, capacity, and many others. Fuel cells have a much smaller up-front price, but the low round-trip efficiency can result in a higher overall operating cost if one is purchasing electricity to produce the hydrogen. Although batteries win on round-trip efficiency, they incur a larger capital cost, one that’s proportional to the amount of storage capacity. With hydrogen, however, increasing storage capacity only involves more (or larger) tanks, not additional fuel cells. According to the National Renewable Energy Laboratory (NREL), for durations above 12 hours, hydrogen has an economic advantage over batteries.

Source: NREL.

In the previous example, Microsoft ran its servers for 48 hours, drawing about 12MWh of energy. In a few years, we can expect the price of Li-ion batteries to drop as low as $100 per kWh, so a 12MWh battery will set you back about $1.2 million. At $50 per kW, a 250kW fuel cell runs $12,500. (That doesn’t take into account the cost of storage tanks or solar arrays, but it’s safe to say that those won’t approach the million-dollar mark.) How many MWh of electricity it will produce depends on the size of the hydrogen storage tanks in the system.

Environmental Impact

There’s no technology that’s entirely benign; the goal of sustainability is to do the least amount of damage possible. Manufacturing “green energy” equipment is often a dirty process, especially when the materials require mining, smelting and other processing. The catalyst in most PEM fuel cells is platinum, a rare, precious metal that’s mined primarily in South Africa, with smaller operations in Russia, Zimbabwe, Canada and the U.S. In addition to its use in fuel cells, platinum is also used in catalytic converters and gasoline refineries. While extracting and processing platinum ore has a high carbon footprint, the metal itself is recyclable, which decreases its negative environmental impact. In fact, recycling platinum has led to lower prices, which is part of the reason that fuel cells are becoming more affordable. Even better, researchers are developing catalysts made from platinum alternatives, such as carbon nanofibers and even biosynthesized materials.

More on Efficiency

Fuel cells produce a lot of heat, so they’re often used in combined heat and power (CHP) or co-production systems, which increases their efficiency since the heat isn’t wasted. During the summer months, the excess heat can also be used to cool an area, using an absorption chiller. That seems like less of an issue in this application, since backup generator fuel cells rarely produce energy, but just as otherwise idle data center UPSs are now being considered for grid support, the fuel cells could be employed in a similar capacity. Maybe fuel cells, instead of batteries, will become the new peaker plants.

Will It Fly?

At this point, Microsoft is still in the exploratory stages of using fuel cells as backup generators. The 250kW system was a proof of concept; the next step is to test a 3MW system, which is equivalent to the diesel-powered generators at the company’s data center sites.

In a larger sense, one reason that the hydrogen economy has been slow to expand is the lack of an infrastructure, which is sort of a chicken-or-egg problem: it’s hard to buy into the technology without an infrastructure, but nobody wants to invest in the infrastructure until people are using it. While the government is providing some funding for green energy, we need more companies in the private sector to explore alternate uses for these energy systems in order to grow a green economy and a sustainable environment. Hydrogen has enormous potential in that regard; how soon it becomes an economically viable technology depends on how many companies are willing to take a chance on it.