The current and future cost of fuel cell micro-CHP systems is of great importance, as the battle to convince consumers to purchase these systems will be heavily swayed by how much it is going to cost them. The price one could expect to pay for the systems currently under development, either now or in the future is however poorly understood, and so must be estimated.
At the start of this project there were few, if any, published prices for fuel cell CHP systems as each machine was individually built and not sold without a tight confidentiality agreement. With the exception of the newly released ENE-FARM systems, it is still challenging to find any manufacturers who can or will openly state how much their systems cost to produce today: the strategic value of knowing competitor’s costs is continuing to keep firms quiet.[108] It is therefore difficult to give a single number for the price of a fuel cell micro-CHP system. As an indication, anywhere from €20,000 to €200,000 is required to acquire one,24 with specific prices depending strongly on the manufacturer and order volume. Trying to estimate what these prices would be in other scenarios (e.g. when mass produced or in the future) therefore poses an even greater challenge.[110]
Two estimation methods are typically employed: bottom-up cost analysis and using learning curves. The former method involves estimating the materials and manufacturing costs for each component of the system, while the latter relies on historical data for prices and sales volumes to extrapolate how they will fall in the future.
3.3.1. Bottom-up Cost Estimates
Even if the present cost of manufacturing systems was widely known, it would not give an indication of how much they would cost to build en-masse once they were fully commercialised. The transition from low volume, highly specialised assembly to automated mass-production lines will bring about enormous reductions in labour intensity and plant utilisation, and thus in manufacturing costs. The cost of producing systems at high-volume is therefore estimated to give an idea of where this bottom-line could be expected to lie.
24 This is indicative of the range of costs quoted to the University of Birmingham’s Fuel Cells Group for CHP systems between 2005 and 2009.
To do this, the fuel cell system is broken down into components, and then into individual materials and production stages. These are often parameterised; for example, the area of electrolyte required is expressed as a function of power density, so that a sensitivity analysis can be performed. The cost of each material, process and component is then estimated from interviewing relevant companies or with industrial cost estimation software, which contains a database of reviewed costs for standard manufacturing goods and processes.
Assumptions about the construction and performance of the system are critical to the results, and estimates for the future costs of specific items (e.g. polymer membranes) have to be speculative as there are no solid foundations on which to base future costs. Estimated costs can therefore vary widely between studies as different assumptions are used.[141]
This is most obvious when comparing studies of fuel cell stacks for automotive use and those for stationary purposes such as micro-CHP. The widely publicised high-volume estimates of as little as €15/kW “are not valid for stationary systems”[141] as their design criteria are too different.25 Automotive stacks are usually 50-100kW, with a focus on high power density and low cost, rather than long lifetime on reformed fuels. As seen in Table 3.3, the estimated costs for mass produced stationary stacks and systems are somewhat higher, however they compare favourably with the estimated economic value of such a system to its owner – based on the savings presented in Table 3.1.[20, 43, 124, 137].
When the ENEFARM system was launched in Japan it helped to answer one of the major questions relating to a sound economic assessment: how would the cost estimates given in Table 3.3 relate to the actual price offered to consumers? A chasm exists between these estimated costs and current prices which are over €20,000. Increasing production volumes and continued research into cheaper and more effective designs will help, but it cannot be expected that such reductions will be made in a short time-scale. A different type of analysis is required to estimate how rapidly costs will fall to these estimated levels, and what prices can be expected to be in the near future.
25 Most estimates in literature lie between €15 and €100/kW for production volumes above 100,000 vehicles per year (e.g. [142, 143]).
Stack cost (for 1kW) BoP cost Details Ref.
P
EMFC
€180-5500 / kW €230 pressurised stack produced at 500,000 units per year, with a Domestic system costs were extrapolated from a 50kW
separate assessment for BoP components. (2000) [144] €630 + 260 / kW ~€3,000 3 to 50kW systems were considered at 10,000 per year volume. Estimates were made with industrial cost estimation software
and information from the US Department of Energy.26 (1999)
[141] €600 / kW €175 / kW €190 + Materials cost for the stack and balance of plant, estimated using empirical formulae relating to capacity.27 (2005) [146]
SO
FC
€550-600 / kW manufacturers, estimated at production volumes around 50,000 Manufacturing costs for 3-10kW systems from six American per year as part of the SECA Phase I project. (2007)
[147- 149] €150-450 / kW conducted with sensitivity analysis.Estimated cost of 5kW residential units, 28 (2004) [150] €350 / kW €625 based on the Fuel Cells Scotland stack. (2006) Estimated materials cost for a 1.3kW system [151]
A
FC
€600 / kW The actual bill for materials required to produce an Elenco V1.1 module, approximately €220 of which was platinum.29 (2003) [152] €220 / kW Astris Powerstack M-250. (2006) Claimed materials cost for the [153] €400-500 / kW Based on a review of reports from DLR, LBST, ZSW, Hoechst & The Royal Institute of Technology in Stockholm. (1992-1994) [56] €130-560 / kW €2-26 / MWh €225 + system, including the cost of soda lime consumption for a COEstimates for high-volume manufacture of a domestic AFC 2
scrubber. (2001) [144]
€200 / kW Projected cost of a Zevco module, which was sold for €1600/kW at the time. (1998) [144]
Table 3.3: A summary of previous bottom-up and materials cost studies for stationary fuel cell stacks and systems. All values are given in 2007 Euros.
3.3.2. Learning and Experience Curves
Several authors have estimated the rate at which fuel cell prices will fall by the use of learning curves. The theory of ‘learning by doing’ proposes that the cost of manufacturing a product decreases with rising production as companies gain the experience required to optimise their process, reduce labour intensity and develop specialised production machinery. This theory gained recognition in the 1960s, and has been widely used to explain the cost reductions seen across numerous technologies and time periods – from the Model T Ford to photovoltaic panels.
The cost reduction achieved each time cumulative output doubles – known as the learning rate – has been 9-27% for most energy related technologies, as shown in three histograms in Figure 3.1. There is “overwhelming empirical support for such a price-experience relationship from all
26 The original report suggested $717 for a 1kW stack, which was modified to be more consistent with the domestic CHP systems in use today. Further details are given in [145] which is provided in Appendix A.
27 The original report suggested $500 for the stack and $700 for the balance of plant. Assumptions for the BoP were modified as in [145].
28 The original report suggested a central estimate of $90/kW, which was modified as in [145]. 29 The original cost was updated to use more recent platinum prices of €32/g.
fields of industrial activities, including the production of equipment that transforms or uses energy”.[154]
Figure 3.1: A comparison of the learning rates for PEMFC fuel cells presented in previous works,[142, 155-161] along with histograms of the observed values for other energy technologies, taken from [162-164].
Two recent reviews of fuel cell cost estimation highlight the extreme difficulty that has been faced in developing such a model for this technology.[110, 155] They conclude that there have simply been too few installations to provide an estimate of the present-day cost, and information about them is kept private by manufacturers, making it “difficult, or even impossible” to calculate a learning rate. It is therefore not surprising that no previous studies have used historic data to develop a learning curve model, and authors have instead estimated the rate at which they will fall in the future.[110]
The learning curve parameters used in previous papers therefore vary significantly due to the different assumptions made. As with the bottom-up cost estimations, fuel cells for vehicle engine replacements have been the main topic of study (e.g. [142, 155-161]). The lower cost per kW of vehicle stacks, and high assumptions for the initial production rate (up to 50,000 vehicles per year) contribute to the low present-day costs presented in these studies, which can be two orders of magnitude lower than those for current domestic systems.
Figure 3.1 summarises the learning rates used in past fuel cell cost estimations, which fall towards the higher end of the range observed for other technologies – averaging 14-28%. Neij argues that modular technologies such as fuel cells have experienced higher learning rates than monolithic products such as turbines, but concedes that rates as high as 30% are rarely observed.[110] Conversely, Schwoon opted for more conventional learning rates of 10-20%, arguing that several components of a fuel cell system (pumps, motors, inverters) are already
well developed within other products, and would not benefit strongly from the early phase of the fuel cells’ own learning curve.[155]