For the commonalities (BOM) among all thin film modules, a set of distinct substrates were chosen: glass, stainless steel, and polyimide. Only those substrates and encapsulation schemes that are already in common use were chosen (e. g., glass/EVA, Tefzel/EVA), and this could be viewed as a limitation of the study (since they are all rather expensive). However, since encapsulation is crucial for reliability, it seemed the proper choice. To first order, any cost breakthrough in encapsulation is likely to benefit all thin films (and x-Si), and could be treated independently. (If individual technologies have uniquely lower cost options for encapsulation, they should try to develop and implement them as fast as possible to gain a competitive advantage over those given here. More likely, they have specific problems that require somewhat more robust encapsulation and incrementally higher cost.)
Then estimates of the rest of the functional aspects of the BOM were added to the three substrate options (see Table 11.1 for a detailed example with glass), except for sales, marketing, management, R&D, warranty, shipping, taxes, and profit. (These are actually added later at the system price level.) Each BOM design had a top and bottom contact (including transparent concutive oxide, TCO); each had bus bars and wires out; each had adhesive where needed. All cost categories were included: materials, equipment and maintenance, labor, facilities. Where materials were the same between designs (e. g., EVA adhesive or Tefzel or TCO, back contact metals) the same assumptions were used throughout, but adjusted if there were differences of thickness or processes. Cell interconnection monolithically or by soldering was also included.
Note that this means that the only aspects of the design that will vary for the nonBOM portion of the thin film device will be between the two contacts (where those contacts include the top TCO and the bottom metal). This approach makes the analysis quite generic, but it does sacrifice some aspects of specificity.
Table 11.1 shows the estimated breakdown of BOM commonalities for the simplest of all thin film designs: a glass-to-glass superstrate with tin oxide TCO on top; and EVA adhesive and glass on the bottom. The second case, last column of Table 11.1, is where a substrate glass is used and the TCO is added before the top piece of glass is added. (The slightly higher cost
|
Table 11.1 Estimated BOM commonalities of glass/TCO/glass module designs; direct manufacturing cost @ 25 MWp/yra, b
|
a For perspective: if a module is 10 % efficient (100 W/m2), $1/m2 is $0.01/Wp; and if annual output is 25 MW, $1/m2 is equivalent to $250 000 of annual costs (about 2 % of total BOM costs). The total BOM (about $60/m2) would then be equivalent to about $15M/yr, or $0.6/Wp. However, at 5 % module efficiency, the BOM by itself would be $1/Wp.
bThe following are not included in Table 11.1: sales, marketing, management, R&D, warranty, shipping, insurance, taxes, and profit.
for this second case reflects the volume advantage of buying glass/TCO from a glassmaker instead of making it in small volumes.) It should be emphasized that all numbers in this chapter are estimates and will likely change over time due to design changes, volume purchases, or innovations. Some of that is built into later tables.
But the classic glass-to-glass module is not the only possible encapsulation design. Tables 11.2 and 11.3 provide a summary of the entire spectrum of BOM combinations at current cost and production levels. As such, they provide a lower cost ‘floor’ for almost all thin films today (since each thin film has one of these BOM); and that floor tends to be in the $60- $75/m2 range, without a single active conversion element being added to the design. This is an important result, since, for example, the long term **DOE goal for thin film modules is under $50/m2, including all aspects. However, recall that these BOM estimates are a snapshot of current costs. They do not include any improved designs or economies of scale. These can
|
Table 11.2 Summary of estimated BOM commonalities at 25 MWp/yr level, near term
|
be expected to be significant. In fact, one may expect BOM costs to drop by about 25-50 % in some future, steady state, high volume scenario, without significant design changes (see Table 11.5, and discussion below). With design changes (such as replacing the back encapsulation with a thin film barrier layer, or with other radical changes that somehow maintain reliability), the reduction could be even greater. However, a stubborn BOM debit of about $20-$40/m2 (perhaps less with design changes), even in the future, is important to include in planning for thin film research. It tends to keep efficiency high on the research priority list.
A few observations about Table 11.2:
• The use of stainless steel or polyimide today engenders the need for a second, bottom encapsulation barrier, in this case EVA//Tefzel (or EVA//glass), which adds significant expense (for comparison, glass does double duty as a substrate and encapsulant).
• However, flexible modules laminated on roofs have significant balance of system (BOS) advantages for that application, which can more than offset this extra module cost. This will be discussed later.
• Future designs to use less expensive substitutes for any of these materials will demand substantial reliability testing.
• However, volume production and on-site manufacturing (e. g., for EVA, Tefzel, glass) would affect these costs positively. To be conservative, these are not taken into account in this study.
BOM commonalities were also developed for nonstandard designs, e. g., for twojunction thin films both as two and four terminal devices. These required different BOM choices for scribing, for contacts between the top and bottom junctions, and for external wiring. A summary is given in Table 11.3.
Observations about Table 11.3:
• The two terminal design is essentially the same as the basic glass/EVA/glass design for a single junction (i. e., this is like some existing two junction a-Si designs – only a small debit for a tunnel junction is added to the BOM);
|
Table 11.3 Summary of estimated BOM commonalities at 25 MWp/yr level, near term, for multijunctions (all glass/EVA/glass design)
|
• The about $15/m2 added BOM cost of the four-terminal design implies about 1 % in efficiency debit (depending on module efficiency) versus a single junction or a two terminal BOM design, before any additional cost for the semiconductors themselves is added. Obviously, these added costs must be offset by greater efficiency to make this choice cost effective versus the two terminal approach or versus a single junction competitor.
• The payoff for multijunctions is in potentially higher efficiencies, which help offset BOM and area related BOS costs.
To find total module cost, designs for the active semiconductor portions (nonBOM) of each thin film technology were developed, including multiple designs (such as in CIS) where warranted by different processing schemes. In other words, physical vapor deposition (PVD) approaches, chemical vapor deposition (CVD) approaches, precursors and selenization approaches – all the recognized ways that thin films such as CIS, a-Si, and CdTe are made – were estimated. However, no effort was made to exactly replicate any company’s specific approach, for obvious and numerous reasons (e. g., it can’t be done; and it shouldn’t be done for confidentiality purposes). So in that sense, even at the most detailed level, there were some limitations on precision. However, every effort was made to make the numbers true to a current best estimate of costs for plants of the sizes given in some steady state (i. e., without first time design costs). For a look at various cell and module designs and process sequences, see: e. g., Wieting, 2005; Basore, 2004; Delahoy et al., 2004; Guha and Yang, 2003; Enzenroth et al., 2004; Powell, 2004, Jansen et al. (2005). However, this is not an all inclusive list of resources, as pointed out in the introduction.
Table 11.4 shows the nonBOM breakdown of one technology (superstrate batch process a-Si/a-Si on TCO/glass) with the various process steps and other inputs (parallel to Table 11.1 for BOM). Notice that it does not include anything for either top or bottom contacts. It is just the ‘difference’ appropriate to this approach.
Table 11.4 shows the categories used to break down each of the technologies, as well as a sense of what cost estimates were made. The resulting $1.56/Wp direct manufacturing cost for a-Si/a-Si modules at 6 % efficiency seems quite reasonable in comparison to sales prices (about $2.25/Wp in some markets) in today’s low volume, high overhead marketplace (at production levels that are not yet at 25MWp/yr). At 6 % module efficiency, 1 $/m2 is 2 c/Wp (@ 85 % yield), so (for example), nonBOM capital equipment and related maintenance contribute only ten times this, or about 20 c/Wp. This is the advantage of the a-Si batch process and some other similarly low-capital cost approaches (e. g., CdTe). It also means the initial capital investment is low (subtracting maintenance): only $1.8/Wp of capacity (or about $45M for 25MWp/yr capacity). Note that at higher efficiencies, these same dollar per square meter costs would yield much lower dollar per watt costs and initial investments. This is the flip side for a-Si batch processing – the efficiencies are low.
As already stated, in the case of dye sensitized, plastic cells, and quantum dots, which are not yet in prototype production, estimates were less secure. In fact, this is a well known problem with all cost projections: numbers that are further from being reduced to practice are fuzziest and prone to the largest mistakes and biases. In this study, some liberalism was used in estimating the costs of the nonBOM portions of these so-called 3G thin films (because such costs are considered their unique strength); but a more moderate liberalism was used regarding their projected efficiencies (since this is their greatest challenge). However, they were all taken to be stable long term. This is quite optimistic. But the choice was made for a simple
|
Table 11.4 Estimated nonBOM (active materials) breakdown for a-Si double junction made on TCO-glass using the batch process (many substrates at once) @ 25 MWp/yr
|
reason: any major instability problem would probably make these technologies totally noncompetitive (except for unique but small niche markets). In summary: the second generation (2G) thin films (CIS, CdTe, a-Si) were treated rather conservatively (because they are more mature and there is more data); BOM was treated rather conservatively, with no assumptions about large economies of scale; 3G thin films were given the ‘benefit of the doubt’ simply to show their potential value. Predictions for 2G thin films should be seen as ‘realistic’; predictions for 3G thin films as ‘optimistic.’
To allow parametric studies of the technologies near term and at several levels of future development and manufacturing capacity, the source spreadsheet was expanded to provide estimates of future developments in thin films. Both BOM commonalities and noncommonalities (nonBOM) were varied with levels of technical maturity and throughput.
In terms of BOM commonalities, an assumed cost reduction of 10 % was applied at each level of increased single plant throughput. The first level of production was assumed to be about 25 MWp; and subsequent levels were increased to 50,200 and 1000 MWp/yr, with concomitant
|
Table 11.5 A summary of the possible cost evolution of BOM commonalities at different throughput/maturity levels ($/m2); BOM reduced 10 % at each larger production level
|
BOM cost reductions. In the end, the final level of BOM costs was reduced by about 25 % from the original value for each BOM design. This is less aggressive than an 80 % learning curve would imply from these volumes, which would have led to a reduction of about 60 %; but BOM volume production is already being achieved through the existing x-Si volume increases, so BOM cost reductions will likely take a more modest slope with time than nonBOM reductions in thin films.
In a recent study (Keshner and Arya, 2004), a bottom up approach was taken with BOM, showing that volume purchases and make-buy decisions could contribute substantially to cost reductions; as could design alternatives. Their glass/TCO BOM costs were 50 % lower than those assumed here for the long term (about 10-20 $/m2). Part of the reason was more aggressive design assumptions (they replaced the back glass with a plastic barrier layer, although this is technically unproven); and they assumed even greater volumes and resulting cost reductions via, e. g., strategies like a front end glass plant at 3 GWp/yr plant size. In other words, the cost of the BOM assumed here should be considered conservative. For perspective: at 10 % efficiency, Table 11.4 is a BOM cost reduction from 55 c/Wp today to about 40 c/Wp; but Keshner and Arya imply about 12 c/Wp. That is a very large difference, and it would drive large differences in system cost for module technologies with different manufacturing cost/efficiency ratios. Future studies focused on BOM might be able to refine these results.
Next, the technology specific (nonBOM) aspects of each approach were looked at, and a timeline was developed using a bottom up approach. Expected technical improvements were delineated and their impacts assigned. Where possible, clear pathways and mechanistic cost reduction assumptions were used. For example, if layer thickness could be reduced from 3 microns to 1 micron for ultimate, practical (efficient) devices, equal improvement increments were chosen for the various production levels. Other projected improvements included faster semiconductor deposition speeds; better materials use; more uniform layers; higher quality layers; wider substrates; and incorporating a higher fraction of today’s best cell efficiencies into future typical commercial modules. In addition, some volume economies were also assumed (e. g., in the capital equipment costs), though not as aggressively as in Keshner and Arya, 2004. These projections were seeded into the spreadsheet at various levels of maturity and throughput for each of the major cases. The results were costs for the nonBOM of the technologies that varied as follows (Table 11.6). Note the spread in projected cost reductions from the ‘25 MWp/yr’ level to the ‘1 GWp/yr’ level. The amount for a-Si is least (and thus perhaps most conservative); CdTe in the middle; and CIS much larger. This relates to the assumed reduction in capital costs for the CIS; and the reduction in semiconductor materials costs (from thinner
|
Table 11.6 Possible evolution of technology specific (nonBOM) costs ($//m2) by technology
|
layers and better utilization) for the CdTe and CIS. However, it also means that the amorphous silicon projections could be viewed as less uncertain.
To get total module costs, these nonBOM costs were added to appropriate BOM commonalities for the different packaging and substrate designs; and then an ES&H cost estimate for in-plant costs (and recycling for CdTe) was added (this was also in Table 11.4). The total module costs in $/m2 are shown in Table 11.7.
Again, the cost reduction for CIS appears more optimistic (as a percentage change) than those for the other thin films, and this can be attributed to projected capital and materials cost reductions, especially for the evaporation approaches. It could also be attributed to the fact that the nonCIS technologies are all more mature and already in production.
But manufacturing cost is only half the story. The other half is module efficiency, and in many cases it has driven a bifurcation in approaches between optimizing manufacturing costs and optimizing efficiency.
Table 11.8 shows a projection of the evolution of commercial module efficiencies. It should be borne in mind that efficiency affects both the module cost and the system cost; it applies at both these levels to get the final value of a module in terms of the end user (system cost, which is given below in Table 11.13). It can also have an impact in terms of specific applications: e. g., rooftop arrays are often area limited, favoring higher efficiencies. Two rows at the end of Table
11.8 show the relative increase in module efficiency from today’s levels; and a comparison between the projected long term module efficiency and the best laboratory cell today.
|
Table 11.7 Possible evolution of total module costs by technology ($//m2)
|
|
Annual output |
CIS/SS (high capital, poor materials) |
CIS/glass (high capital, poor materials) |
CIS/glass (moderate capital, poor materials) |
CdTe |
a-Si/a-Si batch on glass |
a-Si in-line on flexible (SS) |
|
25 MWp |
8.5% |
11 % |
10% |
8.5% |
6% |
7% |
|
50 MWp |
10% |
12% |
11.5% |
10% |
7% |
8% |
|
200 MWp |
11.25% |
14% |
12.75% |
11.5% |
7.5% |
9% |
|
1 GWp |
14% |
16% |
15.5% |
14% |
8.5% |
10.5% |
|
% increase |
65% |
45 % |
55 % |
65 % |
42% |
50% |
|
% of today’s |
72 % (of |
82% |
79 % (of |
85 % (of |
68 % (of |
84 % (of |
|
best cell |
19.5 %) |
19.5%) |
16.5%) |
12.5%) |
12.5%) |
|
aAn up-to-date table of module efficiencies taken from websites is provided in Appendix 11.2. In some cases, numbers differ slightly from those in Table 11.8. |
The technologies highlighted in Tables 11.1-11.8 are those in first time or prototype production. But they were not the only technology options studied. Studying others brings up a whole new level of uncertainty, and estimates for these must be viewed with increased skepticism. Not only is uncertainty greater, but the choices needed to make the technologies cost-effective at all are much riskier technically. For example, are any of the new technologies stable? Since we can’t say, we assume they are, just to get a start at comparison. But if they are not, then these comparisons are unusable and even misleading if used out of context. (To a lesser degree there remain reliability issues in the 2G thin films, too, and even in newly modified x-Si products.) There are numerous favorable assumptions made to make cost estimates of emerging 3G thin films. It is crucial that the reader not compare them directly to more mature technologies without this in mind.
The newer technologies share the BOM of the other technologies. This is a helpful grounding that can result in insights. The first obvious question: can a new technology even come close to being cost competitive with those already examined? If not, why should we try to develop them?
There were basically two categories of alternate technologies: further variations on the classic 2G thin films (CIS, CdTe, and thin-Si) but not yet in pilot production (e. g., CIS precursor inks, microcrystalline Si); and so-called 3G thin films such as dye sensitized cells, plastics, quantum dots, and tandem CIS/CdTe like cells.
Table 11.9 shows a summary of the estimated cost evolution of these less mature approaches, for just the nonBOM portions (the BOM was assumed to be the same as the others, depending on designs). Since all of these will start production later than current options, they were assumed to be introduced at larger throughput levels just to be competitive. However, this may not be their actual development path. Note the very low nonBOM cost estimates for the 3G thin films – this is their presumed strength. Complications from unknown process requirements or future large scale production problems were not included and may raise these estimates to much higher levels. Such complications may also exist for the other technologies, but with a reduced probability (since they are better known and more mature). Thus all these estimates make the key assumption that everything on the cost side will go well for these new technologies.
Table 11.10 shows the efficiency assumptions for these alternate technologies, which are certainly open to debate. Notice the radically greater spread in future module efficiency estimates
|
Table 11.9 Alternative thin film module options, estimated cost evolution of nonBOM only ($/m2)
|
aThe ‘CdTe/CIS’ nomenclature is used to indicate a future, successful combination of high gap CdTe and low gap CIS; however, this is unproven and not optimized at this juncture and should be considered high risk (see Table 11.10).
versus current status. But this is unavoidable, as any new technology will always start out very low; and without great progress, it will simply disappear.
Table 11.11 shows a qualitative assessment of relative risk among the technologies. It is based on the gap between current efficiency levels and the long-term efficiencies needed for success; and on technical challenges to scale up, pilot production, or commercial success, including stability issues. These risks need to be borne in mind when assessing real status and potential. Due to similar challenges, many early thin film options have already fallen by the wayside despite high hopes, and this is the way the higher risk options should be viewed until proven otherwise.
For the long term scenario (1 GWp/yr production), a breakdown of the module manufacturing costs by materials; labor; utilities and rent; maintenance; and capital is given in the
|
Table 11.10 Alternative thin film module options, estimated efficiency (%) evolution
efficiency levels |
Table 11.11 Estimated relative technical risks of the thin film technologies (1 is lowest risk) to reach their ultimate cost goals (at 1 GW/yr production)
 |
 |
Technology Relative risk Comments on major risks
following figures. Note that this is only the active (nonBOM) portion of the cost, and as such must be combined with BOM, module efficiency, ES&H, and final BOS costs to reach system costs. For example, this picture gives the most favorable picture of the 3G thin films.
Note that the utilities are rather high. Even though thin film modules have energy paybacks of about a year, a year outside is about 170kWh/m2 in an average location. At a nickel a kWh, that’s about $8.5/m2 in electricity cost. But this is shared with the BOM (especially its embedded energy) and will be lower with time (especially for thinner layers, where cost can come down more than half). But even 60 kWh (3 $/m2) is a large cost in Figure 11.1. This also shows how low the other nonBOM costs are.
From Figure 11.1, it would appear that batch a-Si has a major advantage over in-line a-Si (from having lower capital costs). However, once all the other factors (BOM, BOS, module efficiency are included; see Table 11.14) this apparent major advantage almost completely disappears. Then when the advantages of flexible modules show up at the system level, the advantage is reversed: the a-Si flexible module is a more competitive product. This pattern occurs over and over in thin films: apparent advantages at one level of cost or efficiency can be misleading if other factors are not included in the judgment. In fact, this is the general give and take throughout PV technologies: cost on one side, efficiency on the other (and stability as a general requirement). Each technology makes its ‘bet’ on its strength; but in the end, a combination of strengths is required; e. g., CdTe may not be the best in every category, but it is nearly the best in all of them, leading to the lowest system costs among the thin films examined (see Table 11.14).
The CIS approaches (Figure 11.2) run the gamut from high cost, high efficiency to low cost, low efficiency, and everything in between. The higher capital cost technologies are in
 |
Figure 11.1 For low risk options, a breakdown of module manufacturing costs ($/m2) for the active junctions layers (nonBOM) for the long term scenario (1 GW/yr).
Figure 11.2 For CIS-alloy options, a breakdown of module manufacturing costs ($/m2) for the active junctions layers (nonBOM) for the long term scenario (1 GW/yr).
 |
 |
pilot production. Others are just demonstrating cells. Although most of the costs for the CIS approaches are higher than those in Figure 11.1, CIS is known for its high efficiencies, thus making the potential of the options in Figure 11.2 attractive.
The thin film x-Si (in Figure 11.3) is betting on the strength of crystalline silicon as a known technology to surpass a-Si in efficiency, though this may be difficult (due to the indirect bandgap of x-Si). Similarly, there are substantial variations in approaches to thin film x-Si; some choose higher temperatures and may incur extra substrate costs; others at lower temperatures might not be able to make as efficient devices. These are the type of cost uncertainties implicit in the less well developed options. The dye sensitized approach is a radically different PV technology that has the potential for low nonBOM, while maintaining efficiency. However, a minor shortfall in the dye cell’s relative efficiency or some special design requirement in the modules would easily consume this apparent cost advantage, even at the module level. Stability is also an issue, since little is known now about actual dye cell modules outdoors.
It is interesting to contemplate how options with very different nonBOM expenses (e. g., the ‘CdTe/CIS’ two-terminal multijunction versus the quantum dot module, which is over six times lower) can be about the same cost at the system level (see Table 11.14). That is, when BOM is added to both, and then the relative efficiency of the multijunction is assumed to be more than double that of the quantum dot module, the system results actually favor the multijunction. Yet without considering the BOM, BOS, and relative efficiencies, one would miss this.
As can be seen from Table 11.16 (below), the high risk alternatives (except one hybrid version that has never been tried) are hard pressed to approach the potential of the simpler, low and moderate risk single junction CdTe and CIS options. This brings into question their value as research paths. Why work on them if they do not even provide an advantage over lower risk choices? Perhaps this is too harsh. Perhaps as part of buildings and avoiding most BOM and BOS, the lower cost options could do well; but of course, such applications are open to the less risky ones, too. In the end, some of the high risk technologies may find a home as special
 |
 |
25 MWp 50 MWp 200 MWp 1 GWp
aBOM, nonBOM, and ES&H are included; sales, marketing, management, R&D, warranty, shipping, taxes, insurance, and profit are not included in these direct manufacturing costs.
aspects of the other technologies, e. g., the quantum dot technology as a low cost bottom cell to scavenge wasted long wavelength photons at minimal cost. Why might this work when the quantum dot design by itself might not? Because very little extra BOM cost is incurred for the two terminal design; whereas by itself, the low efficiency quantum dot technology would have to carry the entire BOM. And the nonBOM cost of the quantum dot cell may be very small.
Combining all the above derived and assumed numbers, it is possible to summarize the resulting evolution in module cost (in dollars per watt) for each technology (Table 11.12) and then rank them by system price (Table 11.14) the final arbiter of PV module value.
The system level comparison requires including both BOS costs and a mark up for all the missing marketing, management, other sundries, and profit.
Balance of system costs vary with application. For this analysis, large systems that contribute to CO2 reduction were chosen as most apt. Two such systems are large, commercial roofs; and ground mounted systems. Table 11.13 shows the BOS assumptions for large, ground mounted systems. A system today that might be considered an example of such designs is the Springerville, AZ, installation managed by Tucson Electric Power (private communication, Hansen, 2005; and Mason, 2004). It is important to note that low module efficiencies incur a large, area related penalty (more modules are needed to make the same output) in ground mounted systems. (We will see later that this is usually, but not universally, true for commercial rooftop systems.)
Tables 11.14-11.15 show the same evolution of assumptions for large, commercial rooftop systems. However in this case, two kinds of modules and designs are assumed: glass modules with racks, and flexible laminates without racks. The difference is that the area related costs for the flexible modules is much lower due to the absence of racks and also simpler set up and installation. The reduced area related costs allow lower efficiency, flexible laminates to still be competitive with x-Si modules on glass. It also reveals that thin films made on glass modules are not as competitive as those made on flexible substrates for this application.
|
Hardware |
Nonhardware (design, prep, install, ship…) |
BOS total |
Indirect: profit & marketing |
O&M c//kWh |
||||
|
Area related |
Power related |
Area related |
Power related |
Area related |
Power related |
Dollar related multiplier |
||
|
25 MW |
60 |
0.4 |
30 |
0.1 |
90 |
0.5 |
25% |
0.3 |
|
50 MW |
50 |
0.35 |
20 |
0.09 |
70 |
0.44 |
20 % |
0.2 |
|
200 MW |
40 |
0.3 |
15 |
0.08 |
55 |
0.38 |
15% |
0.1 |
|
1 GW |
30 |
0.2 |
10 |
0.07 |
40 |
0.27 |
10% |
0.05 |
|
Table 11.14 Assumed BOS cost evolution of large, commercial rooftop systems (glass modules)
|
|
Table 11.15 Assumed BOS cost evolution of large, commercial rooftop systems (flexible laminates)
|
|
Table 11.16 Comparison of thin film system prices for ground mounted, large systems ($//Wp DC) based on the above data and assumptions
|
Table 11.16 compares technologies at the system level in $/Wp for a large, ground mounted system. For this purpose, an estimated price has been developed from the various costs. The price includes everything: module, BOS, sales, marketing, management, R&D, warranty, shipping, taxes, insurance, profit, and O&M. The assumed margin for the systems is reduced with time and size, and becomes quite low, as one might believe that at the desired multi-100s-of – GWp/yr level for the large volume estimates (and that is the point of this analysis), overheads, for example, will be tiny, as they are in other energy commodity industries like coal.
A comment on Table 11.16. Why aren’t systems selling for the low prices seen here? First, of all the options that are lower than x-Si (at $4.59/Wp), only CdTe is in manufacturing. The price of systems is also ‘what the market will bear’. However, the CdTe system price does seem low, given current experience. This may indicate a flaw in the analysis (perhaps the area related BOS is higher than assumed here) or larger margins for these systems than taken into account here. But also recall that no single company’s approach is the basis of these estimates. Also, perhaps existing CdTe manufacturing is not quite as optimized for 25 MWp/yr production as assumed here.
Table 11.18 shows the comparison of thin films for large, commercial rooftop systems, in this case including the BOS advantage of flexible modules.
Some observations about the tables and figures:
• Although CIS and CdTe dominate the lowest, long term costs by about 30 %, inherent issues with indium and tellurium availability mean that thinner cells (about 0.5-1 micron) would help maximize their contribution to the TW Challenge (see next section). It is not clear they can reach the efficiencies of Table 11.8 at reduced thicknesses. The CIS and CdTe cells alone are probably not going to be the only surviving PV options. But they may be the most economical.
|
Table 11.17 Comparison of thin film system prices for ground mounted, large systems ($/Wp DC) with risks and barriers
|
|
Table 11.18 Comparison of thin film system prices for large, commercial rooftop systems ($//Wp DC)
|
• Cadmium telluride has the opportunity to dominate all markets, but is especially attractive for ground mounted systems. For residential roof top systems, especially small ones, x-Si may still be more attractive due to its higher efficiency.
• Long term, CIS is as attractive as CdTe, but it is about a ‘generation’ in factory size behind CdTe and a-Si. This lag may make it hard for CIS to fully realize its potential. Risks are also higher with CIS, and not all key challenges may have been overcome (first time manufacturing at the 25 MWp/yr level does not exist). Thus long term comparisons that seem to show equality with CdTe are not complete without this risk assessment.
• Despite good potential, the CIS/CdTe multijunction may not play a role long term, because it would have about the same system price as the separate single junctions but would use more rare indium and tellurium per output watt (while also increasing manufacturability complexity, offset somewhat by higher efficiencies). Intermediate term, it might find a niche where efficiency outweighs system cost, e. g., on small roofs, but this is not a key market.
• The ‘top cell CdTe/quantum dot bottom cell’ (current matched, two terminal approach) provides some minor, potential cost and efficiency advantages over the single junction CdTe or CIS separately; though counter intuitive, it is a potentially sensible way to scavenge low energy photons if the bottom cell can be added cheaply (e. g., using quantum dots, plastic, or dye cells), without damaging the CdTe top cell. This is an ultrahigh risk, speculative option that has never even been fabricated in the lab.
• Thin film silicon approaches (including both amorphous and nanocrystalline silicon) separate into two categories: those on glass and those that are flexible. The ones on glass all have system prices between traditional x-Si itself and the CIS-CdTe complex. However, near term, the thin Si technologies on glass have a hard time competing because they are not leaders in any category, trailing both x-Si and CdTe. However, flexible thin Si (e. g., amorphous silicon on stainless steel) is competitive at the system level versus x-Si for large, metal roofs – an important market.
|
Table 11.19 Comparison of thin film system prices for large, commercial systems ($//Wp DC) with risks and barriers
|
 |
 |
 |
Figure 11.4 Risk and reward by technology for large, ground mounted systems (1 GWp); the CdTe technology stands out for low cost and low risk.
• Dye cells on glass have many aspects in common with thin silicon, except they seem to have lower capital costs. However, dye cells have not been manufactured, and skepticism remains about their reliability outdoors. Aspects of module design to overcome stability issues may lead to added costs.
• The other 3G thin films (quantum dots, plastics) are hampered by severely low efficiencies. They simply may be too inefficient to ever be usable except for specialty applications indoors. Even if they progress in terms of efficiency, they are so immature that they may completely disappear due to technical risks of scale up and outdoor reliability. To be competitive, these type of thin films need efficiencies almost as high as the others, and proven stability.
• It is unclear if any of the more exotic 3G options have materials availability issues at this point (ruthenium dye is a clear issue, but future designs may eliminate it; this should be studied).
 |
 |
Figure 11.5 Risks and rewards for commercial rooftops (1 GWp).
• For comparison (and a sanity check), an 15.6 % traditional x-Si module costing $1.85/Wp, based on the most aggressive BOS case above, would have a long term system cost of $2.62/Wp – which, given the current leadership of x-Si in PV, means that x-Si will likely be around as a competitor for the entire projected period. Only thin films truly executing the above scenarios might change this. In practice, with expected vastly expanding markets for the foreseeable future, x-Si and thin films will likely share the marketplace.
• For the most part, there are some critical issues in each thin film that could seriously hamper success; and those who are involved will have to tackle them while also maintaining explosive manufacturing growth. There is no certainty this will happen successfully, and some ‘skepticism’ factor should probably be added to the above cost projections (e. g., in comparison with the crystalline silicon technologies, with their lower risk) to reflect this.
• On the other hand, anything that would move the goals of a technology in a much more positive direction (e. g., a major efficiency advance over the stated levels) would also affect the leadership among thin films and in PV. Certainly, the predicted long term efficiencies of the technologies are something that could change and be very important.
• Any new, lower cost packaging designs beyond those assumed in this study would probably reduce the cost of all modules for which they could be used, assuming reliability could be maintained for any of them (but see next bullet).
• Given their lack of competitive advantage in conventional applications, the 3G options should rededicate themselves to much higher efficiency strategies (e. g., creative multijunction structures) and to seeing whether their specific designs (e. g., lower processing temperatures) allow a unique potential for simpler, less costly BOM than other thin films. The hurdle of outdoor reliability also remains very challenging.
• This analysis shows groupings of technologies, with single junction CIS and CdTe looking the best because of their combination of high efficiency and low manufacturing cost. However, within groupings, it would be premature to use the results to decide that one approach is clearly better than another, given the inherent uncertainties and the need for successful execution. Indeed, there is a similarity of long term potential prices if different approaches are well executed – and the message may not be their potential similarity of cost, but the similarity of the hurdles needed to get there.
For an average US solar location like Kansas City, $1/Wp DC is equivalent to about 6 c/kWh (see Appendix 11.1 for levelized energy calculations). Thus the range of costs in Tables 11.16 and 11.18 (about $1.1 for ground mounted; $1.5 for rooftop) implies about 6-9 c/kWh AC PV electricity. Especially for larger systems, it might be expected that the sunnier locations would predominate, and in that case costs would be more favorable. Overall, these costs appear within the range needed to meet the TW Challenge for PV, especially when it is recalled that rather conservative assumptions were made for the BOM aspects of the modules and BOS for systems.
