...

carbon sequestration by the numbers, part 2 of ???

In the first post in this series, I did a bunch of back-of-the-envelope math to try and get a rough answer the the question “What would it take to get 651 billion tons of carbon out of the atmosphere?”  Looking at that question from the perspective of energy costs, the quick answer is “a fuckton” with the specific answer being something like 36 million terawatt-hours, or about 227 years of human energy production at 2019 levels. For this post I wanted to try and figure out what kind of USD investment would be required to build the infrastructure for the “industrial carbon sequestration” that I was talking about in Part 1 of this series.  Next time I’ll talk more specifically about what that infrastructure might look like, but in researching this post I had one fairly specific goal: to determine point estimates for capital expenditure (or capex) and operating expenses (opex) of a typical reasonably modern oil refinery in terms of millions of metric tonnes per year (Mt/yr) of refining output.  With these numbers, one can then scale up appropriately to a particular desired output and get an idea of the dollar cost to build and operate that production infrastructure; however, it should be noted that this isn’t a terribly precise way to get an estimate, as I’m quite confident that the scale of investment required to build this infrastructure will have major effects on the overall cost of the investment.  I’ll discuss this more towards the end. You might be wondering why I chose to look at oil refining as the “reference industry” from which to derive cost estimates.  The quick answer for that is that oil refining (and chemical processing more generally) is quite similar in many respects to the kind of process(es) I’m imagining being used for industrial carbon sequestration. Anyway, without further preamble, the numbers I’ve come up with are as follows: for each Mt/yr of refining capacity, you’ve got to invest about $2.4 billion 2020 USD to build it in the first place (capex), followed by about $120 million 2020 USD per year to run it (opex).  This DOES NOT include the cost of raw materials, which are a huge determinant of the price of the finished product (more on that below) These figures are based mainly on North American (Canada and USA) refining industry figures, particularly [1] and [2] below.  To get the capex figure, I first looked at [1] which discusses, among other sections, the actual and projected costs of several at-the-time new refineries and expansions in Canada.  These projects seem to be intended mainly to process heavy oils and bitumen due to the expansion of tar sand oil extraction in the early 2000′s and provided a range of capital costs from $0.45 - $1.59bn 2009 USD per Mt/yr.  Another bit of information comes from the construction costs of the Shell Pearl GTL plant in Qatar [3], coming in at around $1.86bn 2012 USD per Mt/yr capacity.  Rounding that up to $2.0bn 2009 USD and then factoring in inflation since then (about 21% from 2009 to 2020) gives us the final $2.4bn 2020 USD per Mt/yr figure.  The opex figure is mainly derived from [2], particularly the “U.S. petroleum refining/marketing general operating expenses“ and the “U.S. and foreign petroleum refining statistics“ data from April 2011.  This is one of the areas where my ignorance around financial reporting is showing itself: via three different methods, I came up with three wildly different numbers for opex per Mt/yr without being sure which one is the most reasonable.  For example, [1] gives a figure of $29m 2009 CAD of “investment” required per Mt/yr, while [4], as near as I can figure, gives something closer to $8-16m 2020 USD per Mt/yr. Using [2] gave a baseline range of between $73-$118m 2009 USD per Mt/yr, so I took $100m 2009 USD per Mt/yr as a nice round midpoint then applied the ~1.2x inflation factor to it. A few more interesting statistics shook out of this research; from [1] and [5], we see that about 150-180 people are directly employed for 1 Mt/yr production and from [2] we can assume that the average US refinery produces around 10 Mt/yr - in 2009 the average was 9.3 Mt/yr but the trend is toward fewer, larger, more complex refineries so 10 Mt/yr might even be on the low side at this point. * * * At this point, we’ve nailed down what I consider to be reasonable estimates for capex and opex, but I’ve left out a big, big piece of the cost estimating: how much will the inputs to the process - CO2, H2O, and power - cost? Remember, the whole point of this project is to get carbon out of the atmosphere!  Fortunately, the Carbon Engineering paper I referenced last time [6] gives us an answer for the CO2 part, at least: $0.69bn capex and $26m opex (2018 USD) per Mt/yr. We’ll need about 3.06 Mt of CO2 for every Mt of finished product (see assumptions discussion for why), which adds on a cost of $2.2b 2020 USD per Mt/yr capex and $83m 2020 USD per Mt/yr opex for the CO2.  As for the H2O, while there are currently many sources of water it is certainly in the spirit of this project to avoid using fossil water or surface freshwater if it can be avoided.  Given the size of the oceans, we can assume there’s all the water we need for this project if we’re willing to desalinate it.  No need to boil the ocean, in other words - just a miniscule fraction of it! From [10] we have figures on the Carlsbad desalination plant: 189,000 m3 of water produced per day from a capex of $0.9b 2014 USD and an opex of $58m/yr.  As we need 1.48 Mt H2O per Mt finished product, the overall capex needed is $21m 2020 USD (that’s millions, not billions) and opex is $1.4m per Mt/yr.  Do I hear you saying “damn, that’s cheap AF?” You’re not wrong, as far as getting water goes, but the real problem with the H2O is that getting the hydrogen out of water so that you can put it into hydrocarbons is massively expensive in terms of power consumption.  Basically, the price of the water itself can effectively be ignored in the cost accounting compared to the price of turning it into hydrogen.  We’ll get into THAT next time! Finally, the last input: power!  I haven’t estimated the power consumption of any of these processes yet - this is already getting really long, so I’ll kick that to next time.  Suffice it to say that carbon-free power (solar or wind) is getting cheap as hell, which is great for this project as the final cost of the finished product will be heavily dependent on power costs.  According to [11], grid-scale solar ranges from $32-42 per MWh, and wind can range from $28-54 per MWh.  I believe this figure represents a combination of opex and amortized capex over the expected life of a given installation, so you can just “drop it in” to add energy costs if you can figure out how much power you need.  For example, using the total energy requirement for carbon sequestration from Part 1 (3.6e7 TWh, or 3.6e13 MWh), assuming $40 per MWh energy costs we find that the cost for the electricity required to get all the carbon out of the atmosphere is a paltry $1.44 quadrillion dollars - roughly 53x the current US national debt.  Better start buying bonds, folks! * * * Ok, so, some discussion of my assumptions, as promised.  I conjecture that, should my back-of-the-envelope vision become reality, the capex estimate will prove too high and the opex estimate too low.  Basically, the magnitude of the infrastructure required to make carbon sequestration happen at the scale it needs to happen will almost certainly represent, in its construction and operation, such a large undertaking that the economy of the planet will be fundamentally restructured to support it.  This makes a certain amount of philosophical sense, if nothing else: as long as humanity has had industrial economies, they’ve been based on burning fossil fuels.  What this project proposes is, in essence, unburning all of the fuels we burned to get our economy to this point - a truly radical shift in perspective. More practically, the size of the endeavor will likely mean that economies of scale will have a big impact on costs, specifically lowering them.  For example, if a standard “sequestration plant” can be designed in a way that it can be relatively easily replicated, capital expenses can go down dramatically.  On the other hand, since the end-to-end process I envision does depend on extant-but-not-yet-mass-produced technology, it could well work out that capital costs are initially much higher than anticipated.  Likewise with operating expenses - this project could drive demand for chemical industrial labor so high that operating expenses are much higher than anticipated, while at the same time the ever-decreasing costs of carbon-free power could turn out to have a bigger impact on costs than anything else.  Basically, about all I can say for sure is that once the ball gets rolling on making this happen, the assumptions this analysis is based on will almost certainly be invalidated.  I’ve been aiming towards conservatism in my estimates - when in doubt, plan for the worst - so hopefully we would see this project turn out to be cheaper (and maybe a lot cheaper!) than I anticipate.  Just to give one example: I’m treating each necessary process separately in terms of capex and opex, which means that I am 100% certain I’ve duplicated costs in my estimates - but how much of an impact that has, I can’t yet say. * * *

A final technical note: To get the 3.06 Mt CO2 per Mt product I started with the assumption that we’d want to end up with the average blend of refined products suggested by [9].  I simplified that down to a blend of six products: still gas, hydrocarbon liquid gas, gasoline, jet fuel, diesel, and paraffin.  Each of these products has  characteristic physical properties [7] [8] and represents a range of different hydrocarbon compounds. [7]  One possibly-unsound assumption I made is that each of these categories could be represented using a single “prototype” hydrocarbon molecule.  From that assumption, I determined the carbon and hydrogen composition by mass for a prototypical “tonne” of finished product: about 836 kg of C and 164 kg of H.  These serve as the basis for subsequent calculations. NEXT TIME: Just what the hell am I “envisioning,” exactly? I.e. what are the processes that might be used and how might they all fit together? How much power will be used to run this process? Why is hydrogen so goddamn expensive?!  And so much more! * * * References: [1] Fundamentals of Refining - Canadian Fuels Association 2013 [2] US Energy Information Administration - Refining and Marketing 2011 ref T28 and T29 (Excel spreadsheets) [3] Quora: How much does it cost to build an oil refinery, answer #2 [4] Royal Dutch Shell Q1 2020 - Financial Statements and Operating Info (ref p.13 - Operating Expenses - Oil Products) and Q1 2020 Results (ref p.5 - Oil Products Sales Volumes) [5] Bureau of Labor Statistics - Petroleum and Coal Products Manufacturing [6] A Process for Capturing CO2 from the Atmosphere [7] Chemical and Physical Properties of Refined Petroleum Products (pdf) [8] Fuels - Densities and Specific Volumes [9] Refining Crude Oil - Inputs and Outputs [10] The Cost of Desalination [11] Lazard - Levelized Cost of Energy/Storage 2019