13 March 2005

Modeling the Future -- Solar PV

So, I'm working on putting together a model as to when, and where, Solar PV will become widely used. The EIA has a model. Roughly, they say: nowhere and never. (They actually say that about 4 TWh of electricity will be produced from Solar PV in 2025.) Hmmm... A bunch of really smart people working for the government say we will never use Solar PV to produce significant amounts of electricity. And I'm going to disagree with them. Must be time for you to start reading something else.

The basis of my model is that there are a series of markets. Each market requires Solar PV to have a lower price before they will adopt it. As Solar PV penetrates a market, the price falls based on economies of scale and the learning curve. The learning curve for Solar PV is, roughly, a drop in price of 20% for each doubling of units built.

What are the markets? And what are their sizes?

Off-grid was the first market. It has been less expensive to install Solar PV in remote locations than run power lines for quite some time.

Japan is the first on-grid market. Solar PV is cost effective in Japan today. Of course, their electricity costs are about 3 times higher than the cost of electricity in the US: 24 versus 8 cents per kwh. Poor Japan. They import coal from China, and China just reduced exports to meet their own needs. Japan imports natural gas from Indochina. They just cranked up their nuclear generation, but had to put that on hold to address safety concerns. Japan is a good model for the future: they don't have any domestic fossil fuels and they love Solar PV.

How big is the Japanese market? In Japan, they build around 500,000 new single family houses a year. Put a new 3KW PV system on each of these, and you have a potential residential demand of about 1.5 GW of solar panels per year. Add in the industrial and commercial sectors, and you're up to 4.5GW of solar panels each year.

As a sanity check, the CIA factbook says Japan consumes about 1,000 G kwh of electricity a year. That's about the electricity you'ld get from 1,000 GW of panels. We expect most locales to be able to supply up to 20% of their electricity from Solar before they have to worry about storing enery and time-shifting demand. That's 200 GW of panels. Assuming panels are replaced every 30 years, Japan could support around 6 to 7 GW of panel consumption per year, if the price is right.

To make life easy, we will divide electric consumption for a region in G kwh by 300 to estimate GW of panels the region can consume. This assumes 1.5 kwh per year per watt installed, a 30 year lifetime for panels, and 15% of electricity will be supplied from Solar PV. This should generate extremely conservative estimates as to the size of a market.

Thus, Japan provides a market of 3.3 GW of demand. The PV industry has plans in place to provide this much manufacturing capacity for 2007. We can expect prices to remain at current levels through 2007.

Prices should drop by about 33% after that. That's enough to make Solar PV cost effective in markets where electricity costs $0.16/kwh. Places like Germany, Denmark, the Netherlands, and Hawaii. These markets are around 60% of the size of Japan. Producing about 5.3 GW of demand. That's not really high enough to lower prices rapidly; not really high enough to reach the next set of markets that open up at $0.12/kwh.

2008 will be the critical year for subsidies (of about 10% of the price of panels) to boost demand to about 7 GW. At that level of production, prices will continue to fall through 2009 and Solar PV will become cost effective in markets where electricity costs $0.12/kwh: California, Spain, Brazil, New York. These markets will sustain a demand of 7GW per year without subsidies. But to reach the next market, at $0.10/kwh, demand needs to increase to 14GW per year, or, we have to wait a year or two for steady production to lower prices via the learning curve.

The above is a rather conservative model. And yet, by about 2013, it predicts that the learning curve will have brought prices down to about $0.08/kwh -- competitive with the average price of electricity throughout the US and most of Europe. That's 20GW per year of PV being installed throughout the world; 10GW per year in the US; 10 G kwh of electricity being produced from Solar PV in the US in 2014.

How did we get to estimating 2.5 times as much solar electric production 11 years earlier than the EIA predicts? The EIA uses a simple model of 17% growth of solar electric production per year, even though the on-grid market has been growing at over 40% per year for the past 7 years. Of course, 2004 showed that fixed rates of growth do not make for good models.

But, what about polysilicon feedstock levels? Won't the lack of solar grade silicon keep the market from growing rapidly?

There are two models we can use for that. One model says, sure, we won't have anywhere near enough polysilicon until about 2008. In the meantime, the industry picks up three years of learning. When polysilicon production picks up in 2008, prices start to fall and the industry rapidly expands module production. By 2014, you end up at about the same place.

The second model notices that astute manufacturers have been planning for a polysilicon shortage for years. REC created Solar Grade Silicon a couple of years ago in order to ensure themselves of a steady supply of silicon. SunPower created panels 33% more efficient than the competition. Evergreen invented ribbon growth techniques to reduce consumption of silicon in half or less. And the industry has been steadily developing equipment to manufacture and work with much thinner wafers.

This evidence suggests that the industry has been aware of the coming polysilicon shortage for a long time and have been planning for it. In the face of the shortage, the industry responded by announcing manufacturing capacity increases that will triple production by 2007. It seems obvious to me that the solar PV industry will continue to grow at 40% per year, taking advantage of improved manufacturing techniques that reduce silicon consumption, and increased silicon production.

Cs