Anda di halaman 1dari 35

Thermo Scientific Prima PRO Advanced Process Control

The world leader in serving science


1

Advanced Process Control (APC)

9% 39% 52% Yield Capacity Energy

Polyethylene APC Example


Ethylene Polyethylene

(
HDPE

H C H

H C H

H C H

H C H

H C H

)n
LDPE

If there is no branching (left), it is linear high density polyethylene - HDPE Branching results in low-density polyethylene LDPE (right) The ratio of hydrogen and ethylene in the reactor feed is an important variable for controlling the density of the polymer

Manual regulation comparison of results by GC and MS


Hydrogen is added according to GC readings and operator know-how

A certain percentage of trials are out of specification

Variations are under-estimated by the GC

PMS versus PGC Polyethylene Production


Hydrogen is introduced in an automatically controlled fashion to control the H2/C2 ratio, first using PGC data then control is switched to the process mass spectrometer The process becomes more stable (smoothed kinetic profile)

Value

Value

Use PMS Speed to Replace Multiple PGCs Use PMS Precision to Improve Control 1% improvement in yield of LLDPE worth $13K per day on a typical unit

Thermo Scientific Prima PRO Ethylene Cracking Furnace: Real-Time Optimization

The world leader in serving science


6

Olefin furnace
Hydrocarbon feed pyrolized with steam at 750-900 C in tubular furnace Gas feed (ethane, propane)
C2H6 CH2=CH2 + H2 C3H8 CH3CH=CH2 + H2

To gas analyzer

Liquid feed (typically C10+, can be up to C40)


CnH2n+2 CH2=CH2 +C3H6+C4H8 + CnH2n+ H2

Residence time 0.08-0.25 seconds, pressure 175-240 kPa Effluent product stream quenched by heat exchanger Various components (particularly ethylene) separated out by further processes Need to avoid undesired reactions
Dehydrogenation to acetylene, MAPD Condensation to cyclo-diolefins & aromatics Coking

Olefins process: schematic

Product Slate depends on cracking severity in the pyrolysis furnaces

Ethylene Furnace Control


Control Variables: Burner Firing Rate Flow Distribution Steam to Carbon Ratio Furnace Draft (Damper Position) Steam Drum Level

These are used to control severity & selectivity so that most profitable gas composition is sent to the recovery section

Real-Time Optimization: Problem statement


To optimize olefins unit profitability we need to accommodate day-to-day changes in target product slate by analyzing market conditions Feedstock costs Energy costs Market value of different olefins Stable operation needs minute-by-minute control of feedstock flows and furnace coil temperatures to maintain appropriate cracking severity determined by RTO model Inevitable deposition of coke on furnace tube walls needs to be minimized and predictable Simulation software is needed so that de-coke procedures can be scheduled appropriately

10

Example: Steady State Control Model

where = equilibrium approach

P = Kp

(yH2)(yC2H4)

P = hydrocarbon partial pressure Kp = equilibrium constant for ethane dehydrogenation reaction at coil outlet temperature (COT) y = mole fraction of constituent in coil effluent

(yC2H6)

Requires: COT measurement Hydrogen measurement Ethylene measurement Ethane measurement

11

Olefins plant automation systems

Plant profitability depends on the quality of the analyzers


12

Open-Loop COT correction: Process GC


COT correction (COT bias) = calculated COT actual COT

Open-loop

According to the Technip SPYRO Team GCs are too slow and unreliable For closed-loop control

13

Closed-Loop COT correction: Process MS


Fast, complete & reliable analysis facilitates closed loop control for more accurate severity control

14

Other APC models


Thickness of coke layer determines energy transfer and residence time Both affect cracking severity and therefore product slate
Tube Section

Coking Rate for Light Hydrocarbon feedstocks

Plehiers and Fromen

rC

AC2H4exp [-Ea,C2H4]
RT

AC3H6 exp [-Ea,C3H6]


RT

Coking Rate for Naphtha and Heavy Condensate feedstocks Reyniers et al

rC

= CH2 CCH4

KiCi
i=1

12

Coking rate models require these furnace effluent concentrations

15

Example model requiring extended analysis


1 Kinetic Severity Function (KSF)* = ln 1- Where is the conversion factor of n-pentane KSF = = = = 1.7 2.3 2.7 3.9 maximize propylene maximize butadiene maximize combined olefins maximize ethylene

Real-Time Optimization (RTO) models change severity set-points after taking into account feedstock analysis, feedstock costs, market conditions (or down-stream requirements) and energy prices
* Stone & Webster technology naphtha feedstock

16

Prima PRO v GC Analysis (severity control)


Liquid feed ethylene cracker Hydrogen Helium Coking rate Methane Acetylene Ethylene Ethane Hydrogen sulfide Methyl acetylene Propylene Propane 1,3 Butadiene Isobutene Isobutane 1,3 Cyclopentadiene Isoprene Selectivity Pentene-1 Isopentane N-Pentane Benzene Kinetic Hexadiene Severity Hexene-1 Hexane
17

MS mean 13.4136 0.0011 28.6200 0.3768 27.5646 4.7831 0.0188 0.2928 12.7321 0.3246 2.9312 3.5980 0.3046 0.9627 0.7922 0.5557 0.1951 0.8675 0.7501 0.2226 0.4439 0.2489

GC mean

28.9980 26.9372 4.5889

12.9908 0.3879

MS analyzes 22 components in 30 seconds 3 minute sample interval 1 MS for 5 furnaces GC analyzes 5 components in 3-6 minutes 1 GC per furnace MS precision ~ 0.1% relative GC precision ~ 0.5% relative Additional extended analysis (slow) GCs needed to provide model input data

Traditional process analyzer installation: GC


1 GC per furnace for cracking severity control
(C1 C3 only) 5 minute sampling interval

1 GC per two furnaces for APC model input


H2, C1 C4, C5+ 30 minute sampling interval

Primary control requirement is to optimize product slate and minimize coking due to excessive carbon radicals
18

Alternative analytical installation: MS


1 MS per 10 furnaces for cracking severity control + APC model inputs (H2, C1 C6)
5 minute sample interval

Although one MS can replace all 15 GCs, in practice a 2nd unit is typically provided for redundant operation
2.5 minute sample interval (99% of the time)

19

Benefits of using MS v GC for furnace control


GC Data v MS Data

GC

Online, near real-time analysis permits tight control of cracking severity in order to maximize the yield of target olefins Advanced control models developed using MS data more accurately predict coke deposition rates which minimizes energy consumption (steam and fuel gas) Tight severity control extends the operating hours of each furnace thereby increasing annual production of olefins High reliability and very low maintenance result in much lower total cost of ownership compared to multiple-GC alternative

MS

Ethylene mol%

On a typical ethylene unit (750 ktpa) 1% increase in ethylene yield worth $20k per day

20

Braskem Evaluation Setup for GC versus MS

6 Furnaces chosen for the test: 3, 5, 6, 7,11 & 12 GC Cycle Time = 9 min MS Cycle Time = 3 min
21

Ethylene furnace effluent sample handling


Trace heated Sample Tube Heated Sampling Cabinet
0-1 litre/min adjustable flowmeter
PI

Reflux Sample Conditioner

To Prima PRO RMS

Membrane filter

Process line

0-1 litre/min adjustable flowmeter

Bypass exhaust

22

MS data validation: Braskem Camaari

Significant GC measurement error when rapid changes occur in cracking severity

23

MS data validation: Braskem Camaari

24 hour cracking severity (propylene/ethylene) following simultaneous calibration of MS (brown) and GC (blue)

24

Stability Data (17 days): Braskem Camaari

This test established a 5-day calibration interval for the GC. 17 days is not nearly enough to assess the stability of the mass spectrometer.

25

MS v GC Severity Plots Braskem Camaari

26

MS v GC Severity Plots Braskem Camaari

27

Additional MS option: Monitoring low-level CO


Continuous addition of ppm levels of dimethyl disulfide (DMDS) into the feedstock can result in a reduction of coke formation by > 50% and CO production by >90% Monitoring CO clearly indicates the effectiveness of the cokereduction strategy and can be used to improve the accuracy of the coke-rate predictive models CO is not measurable in low concentrations by MS due to significant spectral overlap with ethylene and other highconcentration hydrocarbons This measurement is best performed by NDIR analyzers

28

External analyzer modification

Allows diverted fast-loop sample to be directed to external CO analyzer

29

External analyzer modification


Allows for otherwise-impossible measurement to be incorporated into MS data by reading analog output from IR analyzer Thermo Scientific GasWorks software presents CO data as though the measurement were made by Prima PRO Alarms on the CO concentration can be configured to indicate excessive CO production (and by implication an unacceptable increase in the coking-rate)

30

Ethylene production cost components


Naphtha $120 per barrel (130% of crude price) or $900 per tonne Cracker operators in Asia typically need a spread of at least $250/tonne between naphtha and ethylene prices just to break even Therefore they require an ethylene price of $1,150 per tonne of ethylene Current price is around $1,000 per tonne. Clearly profitability requires product slate optimization to reduce required spread to something significantly less than $250

31

Typical GC Solution: Cost of Ownership


Total cost of ownership (10-year):
10YearTotalCostofOwnership Cost Quantity Total GC $90,000 9 $810,000 Service $7,000 90 $630,000 Gases $1,000 90 $90,000 $1,530,000

3-minute measurement interval for C1, C2 and C3 only

32

Typical MS Solution: Cost of Ownership


Total cost of ownership (10-year):
10YearTotalCostofOwnership Cost Quantity Total MS $175,000 3 $525,000 Service* $7,000 7.0 $49,000 Gases** 0 0 0 $574,000
*First 3 years included then one service every 3 years per machine ** No fuel or carrier gases required

90 second measurement interval for extended analysis to include hydrogen and speciation to C6
All furnaces connected to all Mass Specs for 100% availability
33

Opportunity Cost
Assumptions Olefins production 700,000 tonnes per year @ $1,000 pt $700M pa Each 1% improvement in yield = $7M pa The improvement in the yield of ethylene and other high-value hydrocarbons is expected to be at least 1% when MS monitoring replaces GC due to the extended analysis, superior precision, stability and availability If we add the 10-year opportunity cost to the TCO10 in the 9-furnace example we get the following cost comparison:

GC $1.5M + $70M $72M MS $0.6M Note that a 3% improvement is more typical


34

Summary
Tight control of cracking severity requires online furnace effluent analysis to maximize profit Furnace effluent analysis cycle time needs to be 3 minutes or less to track process kinetics Prima PRO provides Superior precision (x5) Extended analysis (speciation to C6) Higher availability (100% with redundant installation) More flexible configuration (configured with software rather than hardware) Prima PRO provides rapid ROI and low TCO10 due to 3-year maintenance interval and 3-4 hour PM All Prima PROs are shipped with performance guarantee and 3year parts & labour warranty with travel expenses also covered for the first 12 months, significantly reducing risk of switching to MS
35