The Economist recently highlighted a few promising developments in high-efficiency solar photovoltaic (PV) cells, specifically naming Oxford PV’s tandem cell technology, but the article neglected the punchline of what the impact would be (“Solar’s new power,” May 23, 2020). In fact, it could be quite significant. Some years ago, the cost of a solar plant was dominated by the cost of the panel, but now project costs are spread among a variety of categories. In the industry’s current stage, increasing the yearly power generation of a solar plant can be one of the most substantial ways to reduce cost since this affects the full stack of costs. To get the “$/MWh” to go down, boost the “MWh” in the denominator. Furthermore, there are several remaining major opportunities for boosting power generation for solar plants, including perhaps with Oxford PV’s technology.
Impact of Component Cost Reduction
There is no problem with reducing the cost of each individual aspect of the plant, but each effort only affects part of the cost stack-up. As an example, suppose clever engineers make a breakthrough with solar tracker design and manage to lop off 30% of the tracker cost. This impressive feat would deliver less than a 3% cost reduction for the solar project, as noted in the example U.S. utility-scale project cost breakdown in Table 1 (Structural Balance of System is the tracker).
Boosting Yearly Power Generation to Reduce Cost
The “cost of solar power” generally refers to a solar plant’s required sale price for electricity to provide an adequate return on capital for the project. Setting aside financial factors and the sale price, the two main drivers that affect the project’s cash flow are plant capital cost and yearly revenue, since operations and maintenance costs tend to be small. Revenue is proportional to yearly power generation so a simple strawman metric to indicate higher or lower cost of solar power is capital cost divided by yearly generation of AC power.
The math for this capital cost figure plays out with the rough example numbers in Table 2, which assumes a single-axis tracker project in Phoenix, AZ.
Suppose instead these clever tracker engineers from the last example directed their resources to boosting yearly generation instead of cutting cost (of course, they should do both). Their efforts would increase the plant revenue and therefore the entire plant’s capital cost in terms of ($ / (MWhAC/yr)). Recently, NexTracker did just this with their TrueCapture control system package, which they claim can achieve up to a 6% increase in plant yield. Holding the plant cost constant, this would translate to a 5.7% reduction in cost per yearly generation, as noted in Table 3 (I don’t know NexTracker costs or performance; I’m just working out example numbers.). As an aside, NexTracker’s new control system’s algorithm reduces inter-row shading for non-flat sites and adjusts aiming during cloudy or hazy weather. It would actually deliver the most benefit somewhere much cloudier and hillier than my example location of Phoenix.
A more sophisticated tracker control system apparently delivered double the impact of a 30% tracker cost reduction. By increasing generation, a component supplier improved the cost per generation of the entire cost stack-up rather than just its own pie wedge of the cost breakdown. A counterpoint may be that several tracker startups likely believe more than a 30% cost reduction is possible and that both lower cost and higher output are achievable. It is an “And World” after all.
Avenues for More Generation
Not only is increasing generation a productive way to reduce cost, but there are large remaining opportunities to do so. Five example avenues of varying degrees of difficulty are as follows:
- Increase crystalline silicon panel name-plate efficiency
- Use cells with a higher theoretical maximum efficiency
- Operate panels cooler to increase real-world efficiency
- Generate power from light hitting the back of the panel (bifacial boost)
- Track the sun about two axes
In this post, I will comment on the first two points, which pertain to solar panel improvements and defer to the next blog post points 3, 4, and 5, which pertain to the mounting structure and power plant design.
Approaching the Theoretical Efficiency of Crystalline Silicon Solar Cells
The efficiency of solar panels used in utility-scale plants now exceeds 20%, and there is substantial headroom for continued improvement in efficiency. While the Shockley-Queisser theoretical maximum efficiency for silicon is approximately 33% (see Rühle), according to Andreani et al., a more-practical upper limit is 29%. Furthermore, 26.7% has been achieved at the cell level. There is still substantial room for continuing to increase panel power. If commercial panels inched efficiency up gradually to say, 26% without compromising panel cost on a cost per area basis, the boost in yearly power generation would deliver a 23% overall cost reduction. Table 4 shows a rough comparison.
Solar Cells with Higher Theoretical Efficiency
An alternative way to boost power is to use a multijunction cell design with a higher theoretical efficiency limit. A solar cells passes photons with a longer wavelength and lower energy than its p-n junction band gap. For photons with higher energy than the cell band gap, only the energy equal to the band gap is used to generate electricity, and the rest is wasted to heat. Solar cells can be constructed with more than one junction with higher energy band gap junctions layered above lower energy band gap junctions to use better a wide range of light wavelengths incident upon the cell. The catch is a substantial increase in cost and complexity. Furthermore, considered designs have historically used different cell materials than the world’s 100GW or so of silicon solar cell manufacturing capacity. Multijunction cells have been punished for their high cost and relegated to outer space, in other words, for satellite applications which place a high premium on power density.
Circling back to The Economist article, Oxford PV among others is pursuing a two-junction cell, called a tandem cell, where a perovskite layer is formed on top of a silicon cell to make better use of high-energy photons hitting the cell. The record efficiency for a tandem perovskite/silicon cell was recently set at 29.1%, and the theoretical upper limit is perhaps 40%, according to Andreani et al. The Economist notes that Oxford PV has demonstrated 28% and claims that efficiencies in the mid-thirties are realistic. Critically, this cell design would leverage the silicon solar panel manufacturing ecosystem and incrementally add to the manufacturing process. By doing so, ideally production costs will be low enough for tandem cell panels to be commercially viable here on Earth.
Suppose a 35%-efficient solar panel became a reality, and pretend the cost per panel area remained the same as now (a very aggressive but simple assumption). By similar very rough analysis as above, panel power is boosted 75% versus a 20% efficient panel, and cost per yearly generation would fall by 43%. Even if panel costs are higher, the rest of the cost stack-up remains unchanged with a more powerful panel. If the panel cost doubles per unit area, the overall project cost still falls by 22%.
Some technologies, such combined cycle power plants, are relatively mature, and enormous effort is required to eke out an extra percentage point of efficiency. Not so for solar photovoltaic technology. Every year panel power is a little higher, and the cost of solar power falls a little further. Cutting the cost of individual components adds up but not as fast as boosting the generation of the entire plant.
The next post will cover opportunities to boost yearly power generation from the power plant design such as for designing for bifacial panels and by improved tracking.