How common is solar panel failure?

How often do solar panels fail?
It's a question we have been asked over and over again, although often it's phrased differently.


Is this panel brand better than that one?
Should I buy a more expensive panel?


If you prefer to watch our video on solar panel failure rates??


Thanks to the magnificent knowledge base that is the National Renewable Energy Laboratory (NREL) we can answer that important question with some real life meaningful data.


It would appear that the longer a solar panel sits on your roof, the greater the chance that a defect will occur. No rocket science proficiency required for that deduction, but what IS interesting is that the increased failure rate is quite linear.


By that we mean that there appears to be a 1 in 4,000 chance of a defect in the first ten years, 1 in 2,000 in the second ten years and 1 in 1,000 in the third ten years.


Let's put that another way.


Imagine that there were 200 homes on your street, and you all decided to install a 6.6kW system tomorrow, all with 20 x 330W panels. So that is 200 x 20 = 4,000 solar panels. In the first ten years, one of those panels on one of those 200 homes is likely to develop a defect.


In the next street only 100 families decide to put up their 20 panel solar installation. On their street, after twenty years one panel, on one of those 100 homes is likely to have failed.


Finally, and yes, we are dragging this out a bit, the third street has 50 homes, all putting 20 panels up tomorrow. In thirty years time, one panel on one of those fifty homes is likely to have a defect.


It's really not a high failure rate at all, and as the report mentions, the manufacturing standards seem to be improving, presumably as manufacturers have moved away from assembling and soldering panels by inconsistent humans, and now do it all with robots.

Surely we must be referencing top quality solar panels?

Far from it. The report compiled by NREL is from 54,500 solar installations with just the usual ratio mix of regular, mid range and premium panels.


Premium panels are even less likely to fail

There's no long term hard and fast evidence to support that yet, but it seems highly likely based on what the manufacturers of the premium priced panels claim. SunPower for example say 1 in 37,000 of their panels fail although they don't say over what time frame. REC say 1 in 10,000 fail, again, over an unspecified time frame.


The warranties on the top panels are now 25 years. That's for both the defect warranty and the cell degradation (power) warranty. Regular panels are still usually 12 years, and even though the failure rate is very low, it's not hard to see why the manufacturers of regular panels are not wanting to increase the warranty unless their competitors force them to.


Let's take World No.1 Jinko who make about 25 million panels a year. Their warranty period is 12 years and they do not include the labour cost in that warranty. Let's assume that 1 in 4,000 failure rate holds true during the warranty period, then they can expect that they will have 6,250 warranty claims to pay at about $150 cost each. Over ten/twelve years, that's not too bad, just under $1 million.


However, if they increase the warranty to 25 years and include labour as the top panels do, then not only can they expect something like 15,000 claims on each 25 million panels, but the cost, once labour is included jumps up to about $400 a panel. Now it's $6 million in warranty costs over 25 years. Sure, that's still very small change for a business selling over $3 billion a year, but if your competitors, who for Jinko are Canadian Solar, Trina, JA Solar, Risen and about thirty other huge Chinese firms aren't doing it yet, then why should they?


When REC increased their warranty to 25 years and included labour, I predicted on this website that by the end of 2019 arch rivals Q.Cells in the 'mid-range' panel category would do so too, and lo and behold, November 2019, their Q.Cells Duo panel gets both a name upgrade to Duo + and a nice new shiny 25 year warranty.