FLOWSHEET OPTIMIZATION FOR MULTI-PRODUCT AIR SEPARATION UNITS

Bruce K. Dawson, Scott C. Siegmund, Zhang Yonggui,

Air Products and Chemicals, Inc. The First Baosteel Annual Academic Conference, Shanghai, China, May 27-28 2004 Abstract: An Air Separation Unit (ASU) produces oxygen and nitrogen by the cryogenic distillation of air. The majority of modern ASUs produce oxygen by the pumped LOX process (internal compression process), where the product oxygen is produced at an elevated pressure by boiling high-pressure liquid oxygen against high-pressure air. This flowsheet substitutes a Booster Air Compressor (BAC) for a product oxygen compressor. Some applications require a large quantity of nitrogen at an elevated pressure, as well as high-pressure oxygen. It is possible to produce nitrogen as a low pressure gas from the upper column of the ASU, as a medium pressure gas from the lower column, or as a high-pressure gas by pumping liquid nitrogen and warming it against high-pressure. A turbo expander is typically used to expand air or nitrogen from a higher pressure to a lower pressure to produce refrigeration for the process. Various expander configurations are possible. The best choice depends on the quantity and pressure of the high-pressure gaseous products, as well as whether or not liquid co-production is required.

The machinery selection has a significant effect on the total cost of the ASU, and therefore must be carefully considered when evaluating competing ASU flowsheets. When nitrogen is produced at elevated pressure, the optimum ASU flowsheet may require additional feed air compared to when the nitrogen is produced at low pressure and compressed externally. Minimizing the total number of compression stages often is the minimum capital cost solution. The applicability of vendor standard compression equipment may also be an important factor.

This paper discusses the advantages and disadvantages of co-producing nitrogen in varying quantities via each of the methods listed. Possible turbo expander configurations are considered. Operability advantages and disadvantages of various alternatives are discussed. A recent case study for a large fertilizer project is presented to illustrate the method of evaluating several competing ASU flowsheets. Keywords: Oxygen, Nitrogen, Pumped LOX, Pumped LIN, Optimization 1. Background

For many years the most common air separation unit (ASU) flowsheet produced oxygen via cryogenic distillation of air at a pressure only slightly greater than atmospheric pressure. The Double Column Cycle has been used for over one hundred years to produce oxygen from air. The feed air is compressed to approximately 600 KPa , cooled to near its dew point, and introduced at the bottom of the high-pressure column (HPC). Nitrogen, being more volatile than oxygen, concentrates in the vapor as it rises through the column. The vapor from the top of the column is condensed against boiling liquid oxygen in the reboiler/condenser. The

condensed overhead is divided into a reflux stream that is returned to the HPC, and a reflux stream that is sent to the top of the low-pressure column (LPC). The oxygen-enriched HPC bottoms is sent to an intermediate stage of the LPC. Oxygen concentrates in the bottom of the LPC.

In the Low Pressure GOX Cycle (Figure 1), the product oxygen is taken as a gas from the bottom of the LPC. In this flowsheet, the oxygen product is compressed to the required pressure via a product compressor. This is an energy efficient method of producing oxygen, but oxygen compressors have inherent safety issues, and are therefore more costly, less efficient, and less reliable than air or nitrogen compressors of an equivalent capacity.

W A S T E

L P G A NH P G A N

H P G O XL P G O X

W A S T E

H P G A N

G O X

E x p a n d e rA ir

H P C

L P C

L P G A N

L IN

C L O X

Figure 1 – Low Pressure GOX Cycle

The refrigeration required to compensate for heat leak from the environment, as well as for the temperature difference between the feed air and the product and waste streams leaving the main heat exchanger, is provided by expanding a portion of the feed air into the LPC. The work done by the expander can be recovered as electricity in a generator linked to the expander, or the portion of the feed that is to be expanded can first be compressed to a higher pressure in a compressor mounted on the same shaft as the expander. This compressor and expander combination is commonly referred to as a compander. Using a compander results in a higher pressure ratio across the expander, and lower expander flow to produce the same amount of refrigeration compared to a generator-loaded expander. The optimum choice of expander configuration depends on the total amount of refrigeration required.

If co-product nitrogen is required, it can be taken from either the top of the HPC, or the top of the LPC. Removing nitrogen gas from the HPC reduces the duty of the reboiler/condenser, and reduces the boilup in the LPC. For a given oxygen product flow and purity, there is a minimum boilup required in the bottom of the LPC. The amount of gaseous nitrogen (GAN) withdrawn from the HPC, as well as expander flow bypassing the HPC, affects the oxygen recovery. Lower oxygen recovery implies that more feed air must be compressed to obtain the required oxygen product flow.

Producing nitrogen at low pressure from the LPC generally doesn’t affect the oxygen recovery, but more energy is consumed to compress the nitrogen in a product compressor. Choosing the optimum flowsheet entails balancing the operating cost and capital cost. The number of compressor stages directly influences the capital cost as well as the energy consumption. Compressor vendors have standard frame sizes. Often the lowest compressor cost per KW of compressor power is achieved by maximizing the capacity of a standard frame. 2. Pumped LOX Cycle

Most of the recent ASUs built within the last 5 years have utilized the Pumped LOX Cycle (Figure 2) to produce oxygen at elevated pressure directly from the coldbox, instead of via an oxygen compressor. In this flowsheet oxygen is taken from the bottom of the LPC as a liquid. It is pumped to an elevated pressure and then warmed to ambient temperature against high-pressure air feed. A portion of the feed air from the Main Air Compressor (MAC) is further compressed in a Booster Air Compressor (BAC) to supply this high-pressure air.

For product pressures below the critical pressure, the air pressure must be significantly higher than the oxygen pressure in order to efficiently exchange heat between the air and the oxygen. Figure 3 shows the minimum air pressure as function of the oxygen pressure.

WASTE

LP GANMP GANHP GOX

LAIR

LOX

HPC

LPC

MAC BAC

LAIR

LAIR

Waste

MP AIR 0

1000

2000

3000

4000

5000

6000

7000

0 1000 2000 3000 4000 5000 6000

GOX Pressure (KPa)

Air

Pres

sure

(KPa

)

Figure 2 – Pumped LOX Cycle Figure 3 – Minimum HP Air Pressure vs. GOX Pressure

For oxygen pressures near or above the critical pressure there is more flexibility to vary air pressure. In general higher air pressures give more thermodynamically efficient heat exchange, but practical considerations such as minimizing total number of compression stages may limit the air pressure.

The cold HP Air leaving the main heat exchanger becomes mostly liquid when its pressure is reduced before entering the distillation columns. The reduction in vapor feed air to the bottom of the HPC results in a reduction in LPC reboiler duty. Because the oxygen product leaves the LPC as a liquid instead of a gas, the net boilup to LPC is the same as for the LP GOX cycle. Because less nitrogen is condensed, however, there is less pure nitrogen reflux available for producing pure nitrogen co-product.

3. Pumped LOX/Pumped LIN Cycle

When high-pressure nitrogen product is required, it may be economic to pump liquid nitrogen to an elevated pressure and warm it against high-pressure air instead of compressing it in a product nitrogen compressor. Figure 5 shows the flowsheet for this Pumped LOX/Pumped LIN process. This flowsheet condenses a larger fraction of the total air feed. Therefore it is often more economical to take the compander flow from an intermediate stage of the BAC, and direct the expander discharge to the HPC instead of the LPC. Some

additional power is required for the BAC, but the expander flow does not bypass the HPC. A higher oxygen recovery is possible because the expander flow produces boilup for the LPC.

W a s teH P G A N

H P G O X

L A IR

L O X

H P C

L P C

M A C B A C

L IN

L A IR

W a s te

Figure 5 – Pumped LOX/Pumped LIN Cycle

Nitrogen product compressors don’t have the safety or reliability drawbacks associated with oxygen compressors, but if the required nitrogen pressure is compatible with the air pressure needed to vaporize the oxygen product, a significant cost savings may be realized by eliminating the nitrogen compressor. For GAN pressures greater than 2000 KPa, an air pressure equal to 90% of the nitrogen absolute pressure may be used.

4. Expander Configuration and Additional Liquid Production

For any of the flowsheets presented so far, the optimum choice of expander configuration may depend on the total amount of high-pressure nitrogen product, as well as the total refrigeration requirement. If additional liquid products are required for backing up the gaseous product supply, or for the merchant liquid market, the expander flow may become too high to expand into the LPC. It may be desirable to compress the expander flow to a higher pressure in the BAC in order to increase the pressure ratio, and therefore the work available per unit of expander flow. Designing for a large refrigeration requirement increases the minimum power consumption for the ASU, since machinery turndown is limited. Understanding how the co-product nitrogen demand varies in relation to the oxygen demand, as well as how variable the total refrigeration requirement will be is helpful in evaluating competing process flowsheets.

5. High Nitrogen/Oxygen Requirements – A Case Study

A typical large fertilizer project requires a large amount of high-pressure oxygen, as well as an even larger amount of nitrogen at four different product pressures. Table 1 gives the approximate product flow requirements and pressures for a typical project.

Consideration should be given to

producing at least a portion of the nitrogen from the HP column to reduce the total number of compressor stages. In order to produce oxygen at 4600 KPa, the minimum air pressure from Figure 3 is 6600 KPa. The nitrogen product pressure that can be produced as pumped LIN with this air pressure is:

Table 1 - Product Requirements Product Flow Pressure

NM3/hr KPa GOX 50,000 4600 HP GAN 25,000 8100 MP GAN 28,500 3300 LP GAN1 18,000 710 LP GAN2 20,000 550

0.9reAir Pressu HP reGAN PressuMax = (1)

From Equation (1) a GAN pressure of 7333 KPa is consistent with the minimum air

pressure needed to efficiently vaporize the pumped LOX. A slightly higher air pressure would result in slightly more efficient heat transfer, and would allow the HP GAN to be produced as pumped LIN. Based on the required HP GAN pressure of 8100 KPa, an air pressure of 7290 KPa is needed. The number of BAC stages needed to achieve this discharge pressure can be estimated from the overall pressure ratio for the BAC. The BAC inlet pressure is MAC discharge pressure, less the pressure drop in the front-end air purification system.

Optimizing the overall ASU capital cost vs. energy consumption requires an understanding of the present value for 1 KW of power consumption. Typical values range from $1000 to $3000 per KW. Where power cost is high, a value near the upper end of this range should be used. For projects where power is generated internally, values nearer the lower end of this range are more common.

The optimum MAC discharge pressure ranges from 550 to 700 KPa, depending on the present value for power. Typical values range from 600 to 650 KPa. For the case study we will assume power is relatively inexpensive, and that the BAC suction pressure is 625 KPa. The average compression ratio per stage for the BAC can be estimated for n stages using equation 2.

n1

1

2PP Stage)per Ratio(Avg ⎟⎟

⎠

⎞⎜⎜⎝

⎛= (2)

For P2 of 7290 KPa and P1 of 625 KPa, the Average Ratio per Stage is 1.85 for a four

stage BAC, and 1.63 for a five stage BAC. A value of 1.85 is near the maximum average ratio per stage (the actual pressure ratio will be slightly higher due to intercooler pressure drop). The cost savings for a four stage BAC will usually justify the higher power consumption.

The LP GAN2 product pressure is low enough that it could be supplied directly from the HPC without further compression. One feasible flowsheet is to produce HP GAN as pumped LIN at 8100 KPa , and the LP GAN2 as gas from the HPC. The MP GAN and LP GAN1 could be produced as gas from the LPC and compressed to the required pressures in a product GAN compressor (Case 1). Alternatively, HP GAN and MP GAN could both be produced from the HPC as pumped LIN and both the LP GAN1 and LP GAN2 are produced as gas from the HPC (Case 2). For Case 2 only a single stage GAN compressor for the LP GAN1 is

needed. Producing so much nitrogen from the HPC results in a reduction in oxygen recovery. A larger MAC and BAC are required, but much of the capital cost of the product GAN compressor is eliminated. Incremental capacity in the MAC and BAC may be less expensive than the multi-stage product GAN compressor. We can estimate the overall power for both flowsheets to determine if the compressor cost savings result in a lower overall product cost.

5.1 Estimating Oxygen Recovery vs. HPC GAN production

When a large amount of GAN is produced either as pumped LIN or gas directly from the HPC, there is a reduction in boilup in the LPC. As stated earlier, there is a minimum boilup required to produce a given quantity of oxygen product at a given purity. The required boilup can be estimated based on the L/V in the bottom of the LPC.

Once we know the boilup, the condensing flow from the top of the HPC required to produce this boilup can be calculated based on the latent heat ratio of oxygen at LPC pressure to that of nitrogen at HPC pressure.

The vapor feed to the HPC is slightly less than the vapor flow from the top stage, since the latent heat of nitrogen is slightly lower than for oxygen. For our case study, no liquid products are being produced. For screening the two flowsheets, we will use a value of 1.093 for the HPC Vapor Factor in .

To optimize the design of an ASU, a process engineer must specify the heat exchanger area and pressure drop to balance capital cost against power consumption. More area will result in tighter temperature differences and lower power consumption. For screening competing flowsheets, the BAC flow may be estimated from typical ratios of HP Air to HP GOX or HP GAN. The total air feed flow (the MAC flow) is the sum of the HPC Vapor Feed and the HP Air Flow.

In order to estimate the total power consumption for each flowsheet, we must estimate the expander flow. The expander flow is similar for both alternates, although producing MP GAN at 3300 KPa directly from the coldbox (Case 2) does require slightly more refrigeration than when the MP GAN is compressed in a product GAN compressor. The expander flow for both alternatives is withdrawn after the second stage of the BAC. The estimated compressor flows for each alternate are summarized in Table 2.

Table 2 – Case Study Calculated Flows

Case 1 Case 2 O2 Product Flow NM3/hr 50,000 50,000 Pressure from coldbox KPa 4600 4600 HP GAN NM3/hr Flow NM3/hr 25,000 25,000 Pressure from coldbox KPa 8100 8100 MP GAN NM3/hr Flow NM3/hr 28,500 28,500 Pressure from coldbox KPa 105 3300 LP GAN1 Flow NM3/hr 18,000 18,000 Pressure from coldbox KPa 105 550 LP GAN2 Flow NM3/hr 20,000 20,000 Pressure from coldbox KPa 550 550 Total Pumped LIN NM3/hr 25,000 53,500 HPC GAN Vapor Product NM3/hr 20,000 38,000 V (LPC Boilup) NM3/hr 125,000 125,000 N2 Condensing Flow NM3/hr 173,600 173,600 HPC Vapor Feed NM3/hr 177,100 193,600 HP Air Flow NM3/hr 91,800 119,400 MAC Flow NM3/hr 268,900 313,000 Expander Flow NM3/hr 34,000 40,700 BAC 1st Stage Flow NM3/hr 125,800 160,100 BAC 3rd Stage Flow NM3/hr 91,800 119,400

5.1 Estimating Power

Consumption

For the MAC and BAC the power for each stage can be calculated based on the inlet temperature, pressure ratio, and an average heat capacity using equations 3 and 4. In equation 3,

∆hs is the enthalpy change for an adiabatic compression stage. For either air or nitrogen the heat capacity ratio k = 1.4. The value of C for the MAC is 29.1 J/gmole/°Κ. For the BAC the value of C is 28.5 J/gmole/°Κ. T1 is the inlet temperature in °Κ. We will use a

value of 300 °Κ. The units for ∆hs are J/gmole. Flow in equation 4 has the units NM3/hr. The typical value of the stage efficiency η is 0.85.

∆hs = C x T1 x

⎥⎥⎥

⎦

⎤

⎢⎢⎢

⎣

⎡

−⎟⎟⎠

⎞⎜⎜⎝

⎛−

1PP

1

1

2 kk

(3)

KW/Stage = ηx 22.4x 3600

shx Flow ∆ (4)

Compressor powers for each of the alternatives in the case study are summarized in Table 3.

6.1 Conclusions

Table 3 – Case Study Calculated Power Consumption

Case 1 Case 2

MAC Flow NM3/hr 268,900 313,000 P2/P1 KPa 655/100 655/100 No. of Stages 3 3 Average Ratio per Stage 1.87 1.87 KW / Stage KW 6706 7806 Power KW 20,119 23,419 BAC Stages 1 and 2 Flow NM3/hr 125,800 160,100 P2/P1 KPa 2134/625 2134/625 No. of Stages 2 2 Average Ratio per Stage 1.85 1.85 KW / Stage KW 3079 3918 Power KW 6,157 7,836 BAC Stages 3 and 4 Flow NM3/hr 91,800 119,400 P2/P1 KPa 7290/213

4 7290/2134

No. of Stages 2 2 Average Ratio per Stage 1.85 1.85 KW / Stage KW 2247 2922 Power KW 4,493 5,844 GAN Compressor 1 Flow NM3/hr 48,500 18,000 P2/P1 KPa 710/105 710/550 No. of Stages 3 1 Average Ratio per Stage 1.89 1.29 KW / Stage KW 1232 173 Power KW 3,696 173 GAN Compressor 2 Flow NM3/hr 28,500 0 P2/P1 KPa 3300/710 No. of Stages 3 Average Ratio per Stage 1.67 KW / Stage KW 573 Power KW 1,718 0 Total Compressor Power

36,184 37,272

• Case 2 requires approximately 1090 KW or 3% more power than Case 1, but it replaces a large 6 stage product GAN compressor with a small single stage machine. The cost savings may well be more than the value of the additional power consumption.

• Other combinations of pumped LIN and HPC GAN could also be evaluated. The methodology given here is useful for screening competing flowsheets. The machinery cost is an important consideration in optimizing the total ASU. Adding incremental capacity to the MAC and BAC may reduce the total cost of the machinery.

• The lowest MAC flow is not necessarily the lowest overall cost or the best flowsheet when large quantities of nitrogen product are required.