Should We Worry About Beacon’s Flywheels?

Tom Konrad CFA

How material is the failure of one of Beacon Power’s (BCON) flywheels on July 27?

Last week, I published an article Four Clean Energy Value Stocks I’m Buying Now, including Beacon Power as one of the four.  My rationale for including Beacon was:

Beacon has been operating their first commercial scale 20MW flywheel energy storage plant  since early this year without mishap, achieving full capacity in June.  They are set to begin construction of their second 20MW plant later this year, 54% of the $53 million cost of which will be covered by state and federal grants, making the funding of the plant practical even for a company with a high cost of capital like Beacon.  If both plants continue the relatively trouble-free operation seen so far, that experience will pave the way for less capital-intensive turn-key sales for flywheel energy storage plants worldwide.

I was quickly contacted after my article by two disgruntled Beacon shareholders who informed me that there was indeed a “mishap” at Beacon’s Stephentown plant in late July.  I had missed this incident in my research, which had consisted of a news search, reading management’s discussion and analysis and the financial statements in the quarterly report, and a discussion with a Beacon employee in early July.

I generally prefer not to invest in immature technologies, but was drawn to Beacon because they seemed to have a track record with their flywheels sufficient to mostly eliminate technology risk, given Beacon’s massively depressed price (currently around $1), and the fact that the company is reaching the point of significant revenue ramp, which has the potential of giving the brave investors who buy now returns of many multiples in just a few months or a year. 

Technology risk once again loomed large in my mind.  Given the information I had, I could not decide if the failure represented an isolated incident, or was a harbinger of more to come.  Was this the failure of “one out of 200 flywheels” as management prefers to characterize it, or was it the failure of one of 200 flywheels over the average 3-4 months the flywheels had been in operation?

If we only expect a failure of 1-2 flywheels (0.5% to 1%) over the life of the plant, that should not radically change the economics of the technology.  However, if we expect a failure rate of 1 flywheel every six months or so, that will make a significant difference to the operating costs of a plant that was supposed to be nearly maintenance free. 

Given this unresolved question, and the unpleasant surprise of not hearing about the potential bad news until after the fact (which made me wonder what other bad news I had not heard about) I decided to sell first and ask questions later.  Since it was the day after my article was published, I updated my article discussing the decision.

In Beacon’s defense

Beacon’s position on the failure of the flywheel was that it simply was not material.  The flywheel failed as designed, did not damage the rest of the facility, and its repair was “consistent with the reserves” set aside for operations and maintenance at the Stephentown plant, as Beacon CEO William Capp told me in a phone conversation this week. 

I was never worried about Beacon’s technology being dangerous.  Certainly, you would not want to be standing next to a flywheel when it failed, but such failures are rare even in the most pessimistic scenario I can come up with, and people do not stand around Beacon’s plants for the very good reason that they operate with extremely high voltage electricity.

My question is then, how many more flywheels are likely to fail?  Unlike some commenters, I do not immediately assume that once a flywheel is through a period of “infant mortality” they can be assumed to be safe.  I see two possible sorts of failures: failures due to manufacturing defects and failures due to wear and tear during operation. 

Beacon’s engineers believe that this failure was infant mortality, so if we are to determine how much more infant mortality we are likely to have, we need some idea how long it takes this sort of defect to manifest.  Capp told me on the phone that the particular flywheel in question had been in operation for a “number of months” and that he did not know exactly how long that was.  I’ll assume it was four months, to make math simpler.

Doing the math

It’s very difficult to calculate failure rates from one data point, but with a Bayesian approach we can get some idea of what the failure rate looks like.  I think it’s a fairly safe assumption that a flywheel with a defect is most likely to fail early on, so an exponential distribution is an appropriate probability model.  Since we only have one incident to go by, the best guess for the average lifespan of defective flywheels is the lifespan of the one that broke.  That means that the rate parameter λ will be the inverse of the lifetime of the defective flywheel, and any defective flywheel will have a 63% chance of failing in the first four months, a 86% chance of failure in the first eight months, and 95% chance of failure in the first year.

We also know that the first 30 flywheels which were in operation for one to three years at Tyngsboro are unlikely to have any defects, since any one of those which was defective would have more than a 95% chance of failure by now.  Since the 170 relatively new flywheels at Stephentown have been in service for about 4 months on average, more than half of those with manufacturing defects should have failed by now, meaning that it would be very surprising to see more than one more failure from initial defects, and not seeing any more failures is the more likely possibility.

Hence, if this failure was caused by a manufacturing defect as management believes, they are right that the total failure rate of flywheels from manufacturing defects is immaterial because it is less than 1%.

The other possibility is that the failure was due to wear and tear.  In this case, we would expect the failure rate to be fairly constant over time.  Given the average of two years of operation for the 30 flywheels at Tyngsboro, and the average four months operation of the 170 new flywheels at Stephentown, Beacon has about 120 flywheel-years of experience, during which they have experienced one failure.  So if this was a wear-and-tear malfunction, we would expect 0.6% of flywheels to fail each year, or 12% over the plant’s expected 20 year lifetime.  This was the possibility that scared me into selling shortly after I wrote the article. (At the time, I did not have the numbers I do now, and my off-the-cuff estimate for the failure rate was considerably higher.)

However, we don’t know that the failure arose from wear and tear, and Beacon’s engineers believe that the f
ailure was due to a manufacturing defect.  Given that information, I will assign a 2/3 probability that this was a manufacturing defect, and a 1/3 probability that this was wear-and-tear. 

Given all these assumptions, my estimate of the likely failure rate of flywheels at Stephentown over the plant’s 20 year planned life will be 4%, which is probably low enough to be manageable with the plant’s planned operations and maintenance reserves.

Conclusion

Although my estimates contain more uncertainty than hard data, I now feel that, Beacon management is likely correct that the failure of one or two flywheels in the first year of operation at the Stephentown plant is not material.  Any flywheel failures after the first year of operation would be a much greater cause of concern, as that would lend credence to the possibility that flywheels sometimes fail due to wear and tear, something that would have much greater impact on the cost of operating a flywheel plant.

Hence, I return to my original position that Beacon Power (BCON) is a compelling if highly speculative stock pick at the current $1 price.

DISCLOSURE: Long BCON.

DISCLAIMER: Past performance is not a guarantee or a reliable indicator of future results.  This article contains the current opinions of the author and such opinions are subject to change without notice.  This article has been distributed for informational purposes only. Forecasts, estimates, and certain information contained herein should not be considered as investment advice or a recommendation of any particular security, strategy or investment product.  Information contained herein has been obtained from sources believed to be reliable, but not guaranteed.