Capacity can be a confusing word, with different meanings for different users: it can be a guarantee of minimum design, an estimate of maximum production, a benchmark for performance appraisal, or a valuation tool.

Many users of this site contact me to ask how to interpret capacity data. I’m grateful to a handful of industry leaders who kindly contributed their interpretations of “capacity” to the following discussion, to help me illustrate different approaches to the idea.

Additionally, I’ve just published over 50 years of raw data, courtesy of USGS, which may help for those looking for historical production, consumption, and capacity data. This discussion may help guide its usage.

Issue #1: Unit Conversions

The first and simplest hurdle is unit conversions, between short tons and metric tons (tonnes) [1 short ton = 0.907 metric tons] and between “tons of contained nitrogen” and “product tons.” [1 ton contained nitrogen = 1.216 tons ammonia].

Issue #2: Adjust for Operating Days / Basis Days

The second and far more complex issue is the industry convention of adjusting annual capacity down to reflect the fact that we can reasonably expect a plant to be taken offline for maintenance for a certain period each year.

But how to make this adjustment?

The issue is perfectly captured by the definition of capacity used for data collection by The Fertilizer Institute (TFI), which supplies raw data to many providers of industry statistics, including the US Geological Survey (USGS), and is thus an authoritative source.

DEFINITIONS FOR “ANNUAL CAPACITY”
Maximum annual sustainable capacity of ammonia containing 82% N (product basis).
Units: Thousand short tons of product … Use 340 days of operation per year.
The Fertilizer Institute

To my mind, the problem becomes immediately apparent: is an ammonia plant’s annual capacity equal to its “maximum annual sustainable capacity,” or is it equal to “340 days of operation”?

Although the two might be the same, there is no reason why they should be. The first is assessed, perhaps subjectively, at the level of the individual asset, whereas the second is standardized across every plant at an industry-wide level.

Issue #3: How Many Days in a Year?

I asked one of the major ammonia producers how it defines capacity across each of its different plants:

The plant capacity is your maximum production rate times whatever days you have when you’re in service – if it happens to be 365 because you don’t have a turnaround, it’s times 365.

[One plant] is on a 360 day capacity so we’ll take whatever the ammonia production rate is times 360 – in a turnaround year we’ll use 360 less whatever the turnaround – 345 or 350 or 340.

At [an older plant] we’ll definitely use time for unplanned down time. At [a new plant] we’ll use 360.
CFO of a major US ammonia producer

According to that methodology, the capacity of a plant will change from year to year, depending on the anticipated maintenance work.

This might be useful for internal assessment but, when publishing data for public use, most companies smooth this volatility in plant capacity by reporting an “average annual” capacity, in which the expected outages are averaged not just across one year but across the full turnaround schedule – maybe three or four years. This produces a single, stable figure for capacity – but the plant is likely to produce above capacity in a year with no maintenance, and below capacity in the turnaround year. You might find this unhelpful if, for example, you’re trying to make projections of quarterly supply.

These different operating day assumptions explain some data discrepancies: industry-wide statistics are generally reported using the blanket assumption of a 340 day operating year, but companies publish their own asset-specific data.

I’d suggest that capacity is not a judgemental statistic, just a measure of size. One plant isn’t better because it’s bigger – but it might be better if it runs without interruption. However, by standardizing capacity to an operating year of 340 days, data providers like TFI obscure information about plant reliability. Removing this adjustment would help to compare assets on an apples-to-apples basis.

The good folks at TFI respectfully disagree with me here:

Both pieces of information – “Maximum annual sustainable capacity” AND “the number of days of operation” are necessary to come up with an accurate capacity number. Knowing that we use 340 days of operation allows us and anyone else using the data to back into the annual capacity number if they choose to assume only a week of turnaround or that there will be no downtime at all for a particular plant. Alternatively, we could just ask for maximum annual sustainable capacity; however, then we would not know how many days of operation each respondent is assuming and by adding up the capacity data of all the respondents we would be mixing apples and oranges.
Harry Vroomen, Vice President, Economic Services, The Fertilizer Institute

Given the assumption of 340 operating days, it’s true that we can make further calculations in an attempt to “back into the annual capacity number.” To illustrate, it’s clear that CF Industries thinks that “maximum annual sustainable capacity” at Port Neal should be based on a 350 day year:

PORT NEAL, Annual Ammonia Capacity
data source	metric tons	short tons	operating days
USGS	336,000	370,000	340
CF Industries	345,000	380,000	350 = 380,000/(370,000/340)
Bold text denotes raw data / italics denote computed data. Data sources: USGS Mineral Yearbook, Nitrogen, 2013; and CF Industries 10-K, 2013

But now, which number is correct: is Port Neal’s capacity 370,000 stpy, or 380,000 stpy? Will the plant operate for 340 days or 350 days?

As this site’s members know, I present conflicting capacity data for every plant: a number based on the 340-day standard (provided by USGS), a number published by the company (if available), and my own “adjusted capacity” that assumes a 365-day year. I also publish a fourth data source (when available), which is the capacity defined in a plant’s air permit.

Issue #4: Time

First, data is very often stale by the time it is published: as of April 2016, USGS has most recently published US plant capacity data in its Mineral Yearbook for 2013. Likewise, if a company achieves a small capacity increase in the course of the year, it generally announces this only after the work is done. To make projections of future supply, knowing the capacity given in air permits might help to anticipate future expansions.

Second, any number given for “annual capacity” is a snapshot. By the end of 2016, Port Neal’s annual capacity will have roughly tripled, from 380,000 stpy to 1,230,000 stpy, according to CF Industries’ numbers. The new plant won’t run for the whole of 2016 – they’ve not finished building it yet – but only for a few months. So, even if the new Port Neal plant runs at full capacity when it starts up, the 2016 capacity we’ll see reported by USGS in a few years will overestimate production capability. We have the reverse effect during a year when plants close: end of year capacity data will underestimate production capability.

Third, even after a company has announced minor debottlenecking or expansion work, there’s often a delay incorporating these capacity expansions into industry data. This is one reason why I believe a blanket assumption of 340 operating days is unhelpful: it creates a layer of opacity in the data.

For example, if I wanted to repeat the back-calculation to determine “operating days” for a different plant, I might end up with nonsense: no ammonia plant runs 378 days per year.

DONALDSONVILLE, Annual Ammonia Capacity
data source	metric tons	short tons	operating days
USGS	2,490,000	2,745,000	340
CF Industries	2,767,000	3,050,000	378 = 2,767,000/(2,490,000/340)
Bold text denotes raw data / italics denote computed data. Data sources: USGS Mineral Yearbook, Nitrogen, 2013; and CF Industries 10-K, 2013

The reason for this delay is healthy conservatism in estimates – even if the data lag might be frustrating. This methodology is explained in TFI’s full definition of annual capacity:

Capacity – The maximum production for a plant that is considered by the company to be sustainable for a typical operating year (340 days of operation). This may differ from nameplate capacity for an older plant that has been revamped, because plants often operate above or below design capacity. Thus, for an older plant, capacity may reflect past annual production history if plant capacity has been tested. Capacity for a new plant is the same as nameplate capacity until the plant is revamped or proven to be able to sustain an annual production level above or below design capacity.
The Fertilizer Institute

The data lag comes from the delay caused by needing to know “if plant capacity has been tested,” or if the plant is “proven to be able to sustain” its increased capacity.

If we compare the USGS data from 2013 (published 2015) with older CF Industries’ data from 2011 (published 2012), we can back-calculate a more realistic 353 operating days at Donaldsonville.

DONALDSONVILLE, Annual Ammonia Capacity
data source	metric tons	short tons	operating days
USGS	2,490,000	2,745,000	340
CF Industries	2,585,000	2,850,000	353 = 2,585,000/(2,490,000/340)
Bold text denotes raw data / italics denote computed data. Data sources: USGS Mineral Yearbook, Nitrogen, 2013; and CF Industries 10-K, 2011

In the last few years, Donaldsonville has quietly expanded its annual capacity by 220,000 short tons, none of which is reflected yet in the USGS data (these are “minor” expansions on existing lines, not the new plants).

I describe these expansions as “minor” but, in a good year, this incremental 220,000 stpy expansion means additional product worth maybe $100 million per year. If we don’t adjust for operating days, the discrepancy between the most recent CF Industries data and the most recent USGS data is even bigger, at 325,000 stpy.

DONALDSONVILLE, Annual Ammonia Capacity (short tons per year)
	2011	2012	2013	2014	2015
USGS	2,745,000	2,745,000	2,745,000
CF Industries	2,850,000	2,950,000	3,050,000	3,050,000	3,070,000
Bold text denotes raw data / italics denote computed data. Data sources: USGS Mineral Yearbook, Nitrogen, 2011, 2012, 2013 (most recent); and CF Industries 10-Ks: 2011, 2012, 2013, 2014, 2015

What is Nameplate Capacity?

To better understand annual capacity, I asked a technology licensor how it defines nameplate capacity:

Plant capacity, also called “nameplate capacity,” can be defined, respectively is understood within thyssenkrupp (formerly known as Uhde), as such capacity on which the plant has been designed … under consideration of a defined product quality and ambient conditions. It could also be considered as the 100% plant capacity case. Licensor grants the license for such defined plant capacity and typically has to achieve such capacity during performance testing of the plant. In order to fulfill the nameplate capacity, Owner has to ensure that the provision of feedstock, utilities and climatic conditions are within the specified ranges.

In case the plant does not achieve the design capacity (<100%) although the Owner fulfilled its obligation, such underperformance of the plant is contractually handled by Liquidated Damages (make-good clauses). On the other hand (>100%), it is in nature of the process as well as in nature of equipment design … that a certain extra capacity is achievable without plant modification. However, such extra capacity is typically quite small.
Tobias Birwe, Head of Sales, Ammonia and Urea Division, thyssenkrupp Industrial Solutions

In other words:

Capacity refers to the guaranteed minimum performance by the licensor. There will be a tolerance and conditions that define that guarantee – guarantee documents are usually several pages long, and include quality of the ammonia, delivery conditions, and are based on quality of feedstock, water, ambient temperature, etc, etc. Guarantees typically include emissions, start-up date, and things like that, too.
Steve Lancaster, Senior Technical Manager, Reformer Technology, IHI E&C International Corp

Which is to say nameplate capacity is a performance guarantee … and is therefore a minimum quantity, not a maximum quantity.

Given Nameplate Capacity, what is Capacity?

If nameplate capacity is a minimum, we’re left wondering what the maximum might be.

Capacity would not be an engineering design maximum limit. There would likely be a margin in the guarantee value that the licensor includes to make sure that he meets his guarantees. It could be a percentage or a set margin on capacity, maybe 50 t/d extra.

The way that licensors work, they will set the process conditions, temperatures, pressure, flowrates, etc, that are required to meet their guarantees. They will size whatever equipment is in their scope of work. Then, the engineering company will design the equipment to ensure they meet those requirements, and they may or may not have any margin on the design, depending on what their guarantees are. So, there may be some additional actual capacity in the plant above the guaranteed design basis, particularly when a plant first starts up, the equipment is clean, and the catalyst is new.

In the old days, the calculations weren’t as accurate, and especially in the 60s and 70s when Kellogg was cranking out duplicate plants, there weren’t a lot of different designs – standard sized equipment was used on multiple plants. As a result, those plants tended to have additional margin above their stated 1000 STPD capacity; most of those plants have been revamped to 1500 STPD or even to 1800 STPD or higher, and some of the equipment is still the same as original.

By contrast, today’s engineers have much more accurate calculation tools, and the performance of equipment is better understood, so the modern generation of plants have much less “fat” in them, and tend to operate closer to their design capacity. That extra “free” capacity above nameplate is money that the engineer has left on the table, so he will design the plant to much closer margins than the old days. Every plant today is designed for the specific requirements, based on the local gas conditions, local weather, etc.
Steve Lancaster, Senior Technical Manager, Reformer Technology, IHI E&C International Corp

Technology licensors and engineers define Nameplate Capacity in daily terms for a very pragmatic reason: guaranteed performance must be provable.

Our licensor’s or contractor’s definition of capacity is given in terms of tons per day or similar, which is owed to the fact that it has to be verified in a test run of limited length.

A plant owner’s or investor’s definition might be defined in tons per year and can factor in also expected downtime, but this could not be a guaranteed case because of such unknown factors.
Klaus Noelker, Head of Process Department, Ammonia and Urea Division, thyssenkrupp Industrial Solutions

How to convert Daily Capacity to Annual Capacity?

Nonetheless, annual capacity is often the more useful number so, while the TFI resolutely adjusts everything to a blanket 340 day operating year, other standards were and will be used:

A large percentage of the new plants that are close to starting up now are 2200 MTPD plants. At 365 days per year, that’s 803,000 MTPY. Of course, that rate can’t be maintained indefinitely, as plants require regular maintenance. In the ‘old days,’ we commonly saw 330 days per year used as the basis, allowing you basically one month per year for outages, upsets, etc. These days, plants are typically planning on 3-4 years plus between turnarounds – not always achieving it – which gets you closer to the 800,000 MTPY capacity.
Steve Lancaster, Senior Technical Manager, Reformer Technology, IHI E&C International Corp

Unfortunately, some companies don’t even publish capacity data, and very few make their assumptions regarding “operating days” (or “stream days”) transparent, which is why, despite its limitations, I rely so heavily on the data provided by USGS. For one thing, it’s the only publicly available dataset that spans fifty years with consistent assumptions.

Download this data

Given everything I’ve written above, if the USGS dataset contains too many assumptions for your taste, perhaps the only way to judge capacity precisely is to assess each plant independently, which is exactly what some industry consultants do.

1. Our basic definition of capacity starts with equipment in place and the effective capacity thereof on a stream day basis.

2. That effective daily capacity will, absent other information, start with design. In most cases, however, we’ll adjust this measure according to what is reported/we know about upgrading/experience improvements, etc. Ergo, for the new “standard” 2200 mtpd designs, barring other information we’ll wind up giving it the benefit of a 5+-% expected increase.

3. We annualize resultant daily values typically on assumed regional experience bases, e.g., for North America we typically use 350-355 stream days per calendar year, for much of the rest of the world we use 330 stream days.

4. Ergo the bases we use for capacity only. The relationship of actual output to these data then shows up in our system in operating rates, and effective operating rates (usually) below 100% can be/are for the range of “usual suspects,” e.g., planned maintenance, unplanned equipment outages, feedstock interruptions, weak markets, holidays, war, et al.
Highly respected ammonia industry consultant

Capacity Utilization Rate

The point here is that any measure of “capacity” is by itself meaningless, if what you want to know is how much ammonia a plant will actually produce. The “capacity utilization rate” (or “operating rate”) links annual capacity to annual production.

A lot of people think about onstream rates, but I think that most people, outside the industry, think that it means you were running on a certain day. But the fact that the plant is running at a turndown rate of 50% doesn’t change the onstream rate – so the utilization rate is more [informative]. I care about how much product we’re producing compared to how much we could be. That’s going to come to capacity utilization.
CFO of a major US ammonia producer

USGS also publishes US ammonia production data …

Download this data

… which allows us to plot an approximate capacity utilization rate, simply by comparing annual capacity to annual production.

Download this data

However, to correct for timing issues (plants operational at year end, but not producing all year), and remove the distorting impact of idle plants, USGS recently started to provide an independent calculation of capacity utilization rates, which is probably more reliable, and makes very interesting reading.

Download this data

If there’s a simple conclusion to draw from that chart, it is that capacity utilization is highly flexible. The US used to operate at or above 100% capacity utilization rate in the 1990s – dipping sharply down to just 56% in 2001 and 2003 – and currently sits near 80% (according to the most recent USGS summary data, for 2014 and 2015, not included in the chart). In any case, annual capacity by itself is clearly not a reliable predictor of annual production.

How does this site calculate “Adjusted Capacity”?

The alternate source of “capacity” data that I mentioned above, the company’s air permit applications, is valuable to me because it is no longer a minimum, but a maximum. Permittees want to ensure they remain within their regulated emission limits, even if they run their plants above nameplate capacity. Often the permitted capacity also includes leeway for efficiency improvements or debottlenecks: owners apply for permits above their current production rate so that they don’t need to go through the expensive process of applying for a revised permit sooner than necessary.

I believe that this maximum capacity, indicating the most that an owner believes could be squeezed out of its asset in the medium-term, is the best measurement of “capacity” because – to me – capacity is a word that should convey a maximum.

This leads to many data discrepancies between the figures I present in my “Adjusted Capacity” for a plant, the figures announced by the company, and the figures published by statistical providers like USGS.

Compared to other sources, my “Adjusted Capacity” numbers are biased higher, and thus utilization rates are biased lower, on two counts:

To begin with, they’re based on air permits, where such information is available. I believe this represents the best available assessment of the maximum production potential of an ammonia plant. I admit this is not a flawless methodology, but this is not a transparent industry and there’s no perfect information.

Moreover, I count a year as having 365 days, regardless of a plant’s turnaround schedule. I do this for two reasons:

First, a 365 day year helps to compare assets on an apples-to-apples basis. (Note, TFI describes this as apples-to-oranges. Pick your own fruit.) Capacity is not a judgemental statistic, just a measure of size. Capacity utilization rate is a highly judgemental statistic, conveying success or failure. I judge a plant that performs poorly against a plant that performs well based on the capacity utilization rate of each, not the capacity. Using a 365 day year, I can compare one asset that requires 35 days of maintenance per year with another asset that requires 5 days of maintenance per year. I don’t think it’s useful to say that they both have a 100% capacity utilization rate – though that’s what would happen using 330 or 360 day years. The two plants may be identical in size, on a daily capacity basis, but one is quite simply a better asset than the other – due, perhaps, to age, employee skill, or management’s willingness to invest in maintenance capex.

Second, I established this website with an interest in the next generation of ammonia synthesis technologies. I anticipate that new methods of producing ammonia will become commercialized within years rather than decades, and I see no reason why these new technologies will necessarily be limited by the same turnaround schedule required by the century-old Haber-Bosch process. If a new technology can run 365 days per year, the same performance measurement should be used for an incumbent technology. (I acknowledge many people’s skepticism of new production technologies – that’s a different discussion.)