REN: Solar Encouragement

2021.06.12 – Ian Page

There appear to be two good rules of thumb…

1. Everyone underestimates an exponential

2. Models of the future model the past.

I've blathered about the first a lot but the second, which is basically an article of faith in technology innovation, hit me really strongly while reading a Joule article on solar.

Joule 5 1041-1056 May 2021 (many authors). I shall pick out a lot of points from this excellent paper.

I've commented frequently that the IEA every year has to upgrade its 2050 estimates of the amount of renewables- i.e. it fails to model the exponentials and relies too heavily on what is currently known. Even Bloomberg NEF which does get exponentials, raises its estimates each year.

The paper shows the 2050 projections of a large number of authoritative models (i.e., those that the IEA the IPCC and the EU use) for the optimum balance of solar and wind and points out that the vast majority of a set reviewed use a solar capital cost of 1 euro per watt in 2050, when the AVERAGE in 2019 was 0.995, and the best was 0.618. This essentially invalidates any conclusions or decisions based on them. A set of models called PRIMES used by the EU also use too high a figure. As a result, official estimates of the amount of solar in 2050 vary from IPCC 5 -12.5 petawatt hours per year, while industry sources and researchers based on solar industry roadmaps estimate 41-96 pwh/y. The difference is a factor of an order of magnitude and obviously matters if policies are being based on expectations of only a tenth of what could well be there. For example, Hydrogen electrolysis or Direct Air capture of CO2 would be impossible or really easy depending on the different assumptions behind such estimates.

Solar module learning curves are revealing.   23% cost reduction per doubling since 1976, since 2007 40%.  This is from technology efficiencies, economies of scale, automation, standardization, and a huge body of research into the materials. The latest acceleration has been China standardizing manufacturing tools reducing capital costs as they have gone all in to their policy of owning the solar industry. I have also commented that compared with wind where each turbine has been an engineering project until recent floating turbines, solar is massively standardized, mass produced, and can be rolled out with little engineering expertise.

Non module costs have also reduced LCOE with balance of system costs, cheaper financing, reliability improvements, longer lifetimes, and policy support.

There is also a raft of other innovations I've referred to in the past such as moving contacts from the front to the back raising performance from 20% to 23.5% by 2030 (this is a 15% decrease in costs per watt so it’s not insignificant), Other roadmaps offer 24% by 2030. Silicon/perovskite has achieved 29% and 30% commercial production is anticipated this decade.

Other system innovations have focused on decreasing system costs in inverters and grid connection (chopping peaks off to reduce capacity investment) and increasing value by delivering more solar in the morning and afternoons rather than noon with single axis systems, and by making the inverters useful at night! Bifacials which accept light from both sides also add around 5% output for low additional cost.  AI and other approaches are also reducing operational costs.

I may pick up more of the points in this paper later, but I thought the main takeaway is that there will be many times more solar than published models predict, it will be cheaper than they predict, and that the issue will not be cost or land area but integration with applications, users, grids, and storage

Even though grids and storage look like problems, storage has its own learning curve, off grids may remove much of the grid problem, and grids might even learn how to move faster - it's hard to believe that systems set up 50 years ago, cannot access a massive learning curve. An analogy is media delivery and mobile telecom services, both of which have moved to software driven edge systems where innovation occurs rapidly, allowing the older parts of the system to move more slowly.