Вы находитесь на странице: 1из 8

Predicting middle distillate properties

Development of on-stream correlation models to predict cloud point, flash point and
freezing point and increase production of on-spec jet fuel using a single analyser

MADI ASIRI
Saudi Aramco Total Refining and Petrochemical Company (SATORP)

W
e cannot opti- 59 000 b/d and 97% true net Our first assumption is
mise what we conversion; and a single-stage that, since the internal chem-
cannot measure once-through Isocracking ical structure of middle dis-
continuously and we cannot unit (SSOT) with a name plate tillate shows minor structural
increase refinery margin with- capacity of 59 000 b/d and 50% changes over the catalyst cycle
out continuous process optimi- true net conversion (see Figure life, we can assume that both
sation. This study aims to find 1). cloud point and freezing point
a simple and cost-effective way, Cold flow properties, cloud are functions of the amount of
with no capital investment, to point and freezing point are heavy hydrocarbons in the mid-
measure and predict contin- controlled by the reaction sec- dle distillate; the heavy part of
uously the physical proper- tion. In particular, the amount the middle distillate will impact
ties of middle distillate streams of n-paraffins in the reactor both cloud point and freezing
without the need for additional effluent, which has the big- point. However, since D86 data
on-stream analysers. The first gest impact on middle distillate can give us the required infor-
part of the study focuses only cold flow properties, is con- mation about the middle dis-
on hydrocracking middle distil- trolled by severity of operation tillate streams, how heavy or
late products (diesel and kero- in the reaction section (WABT, light they are and how tight
sene) while the second part will H2PP, LHSV and catalyst type). or wild is the middle distil-
deal with the other middle dis- However, in a hydrocracking late cut, there should be a rela-
tillate streams (CDU, DCU, and process unit, as part of the cat- tionship with the distillation
so on). The model developed alyst selection and catalyst per- data (ASTM D86) and with
in this study can be integrated formance guarantee, yields and both cloud point and freez-
with a DCS (or APC) system product quality must be guar- ing point in the hydrocracking
to minimise product giveaway anteed over the life cycle of the unit. Hence, both cloud point
and maximise the production selected catalyst. and freezing point can be con-
rate of middle distillate within In other words, the chemical trolled in day-to-day operations
final specification limits. structure of the reactor effluent by adjusting the fractionation
Based on a deep understand- (paraffins, iso-paraffins, naph- section’s operating conditions,
ing of hydrocracking process thenes and aromatics) over the keeping in mind that opera-
operation, catalyst function and cycle life of the catalyst must tions in the reaction section
reaction chemistry, we started show absolutely minor changes are almost steady and not fre-
our work with two differing to ensure product quality from quently changed. Usually, the
hydrocracking units: a two- start of run conditions to end of reaction section’s severity is set
stage Isocracking unit (TSREC) run conditions for the selected to meet the nitrogen slip of the
with a name plate capacity of catalyst. pretreat reactor and the global

www.digitalrefining.com/article/1002285 PTQ Q1 2019 1


Make-up H2 10 hydrogen
Make-up
from offpilot compression section
5
0 H2S absorber /
VGO and Two-stage Isocracking Recycle gas compressor
HCGO feed
−5 reactor

CP, ºC
CLPS vapour
−10 HLPS/CLPS Unstabilised to offpilot PSA
−15 Lab Modelliquid naphtha
3wt% UCO bleed Sour LPG to
from MHC−20
unit Recycle gas Fractionation LPG treating unit
separation section section Naphtha
0 10 20 30 40 50 60 70 80
stabilisation 90 naphtha
Light
section
Time, days Heavy naphtha
Fractionation bottoms
Kerosene
product
Diesel
product
Unconverted oil
to offpilot FCC
Make-up H2 Make-up hydrogen
from offpilot compression section

H2S absorber /
VGO and Single-stage Recycle gas compressor
HCGO feed Isocracking reactor
CLPS vapour
HLPS/CLPS to offpilot PSA
liquid
Recycle gas Fractionation Sour offgas to offpilot
separation section section amine absorber
Unstabilised naphtha to
naphtha stabilisation section
Parameters TSREC SSOT in DHC unit
Operating capacity, BPSD 65 000 65 000 Kerosene
LHSV reactor no. (R1/R2/R3), 1/h 1.5/1.45/1.45 1.5/1.45/NA product
H2PP @ reactor inlet, Barg 155 155 Diesel
H2/oil ratio (1st stage/ 2nd stage), Nm3/Sm3 840/680 840/NA product
CATs (WABTs) reactor no. (R1/R2/R3), °C 370/385/350 365/380 Unconverted oil
Net conversion, % 95 55 to offpilot FCC
2nd stage PPC, % 50-55 NA 3 wt% bleed unconverted oil
2nd stage RCP, °C 375 NA to DHC unit
Hydrocrackers average operating conditions during model data collection

Figure 1 Block flow diagram of two hydrocrackers modelled for the study

conversion of the cracking reac- freezing point. A total of eight relation models will add more
tors during the catalyst life correlations were developed for value to process optimisa-
cycle with normal acceptable four product streams in two dif- tions and product quality con-
deactivation rate. ferent hydrocrackers within the trol. Instead of waiting for lab
On the other hand, flash point refinery. These correlations are: results once a day (or in some
is a function of the amount of • Flash point and freezing point cases twice weekly), the oper-
light hydrocarbons (light ends) for kerosene in both TSREC and ations team will have continu-
in the middle distillate, which SSOT ous data and information about
can be controlled in the hydro- • Flash point and cloud point cloud point, flash point and
cracking fractionation sec- for diesel in both TSREC and freezing point to enable them to
tion by adjusting the operating SSOT. take immediate action and not
conditions. These models are tools to wait for the next day’s results
Based on all of the above allow for process optimisation to see the impact of today’s
assumptions and information, to maximise the middle distil- actions on product quality.
linear models were developed late production rate and mini- Cloud point is the lowest tem-
to predict the physical proper- mise product giveaway within perature at which wax crys-
ties of hydrocracking middle product specifications by con- tals begin to form by gradual
distillate. The target properties tinuous measurement at the cooling under standard condi-
are cloud point, flash point and operators’ panels. These cor- tions. However, this parameter

2 PTQ Q1 2019 www.digitalrefining.com/article/1002285


Standard error and model validation range

Models TSREC SSOT TSREC Kerosene SSOT Kerosene TSREC SSOT TSREC SSOT
Kerosene FP Kerosene FP Freezing point Freezing point Diesel FP Diesel FP Diesel CP Diesel CP
Standard error 0.82 1.21 1.21 1.7 1.7 1.4
Validate Range MAX T10%=203 T70%=234 T70%=230 T10%=294 T70%=351 T70%=346
T95%=270 T95%=252 T95%=373 T95%=375
Validate Range MIN T10%=167 T70%=197 T70%=192 T10%=248 T70%=303 T70%=316
T95%=215 T95%=212 T95%=332 T95%=342

Table 1

is an important property of the Objective • Maximise the middle distil-


fuel since the presence of solidi- The aim of the study is to late production rate.
fied waxes can block filters and develop on-stream optimi-
negatively impact engine per- sation tools (models) to ena- Data analysis and results
formance. As the molecular ble refiners to measure Building any model for mid-
weight rises, the cloud point is physical properties in hydro- dle distillates to measure addi-
also incremented at low temper- cracking units continuously, tional physical proprieties
ature conditions. Furthermore, and to have the opportunity demands a deep understanding
cloud point is a measure of par- for continuous process opti- of process operations and what
affin content in wax form, which misation by utilising a dis- impact dependent and inde-
is why no cloud point data are tillation on-stream analyser pendent variables have on the
reported for light cuts like naph- (ASTM-D86). Most refiners target properties. For instance,
tha or gasoline. depend on lab results to meas- in studying flash point the
Freezing point is the temper- ure these additional proper- most important variable is the
ature at which a liquid changes ties, a very slow process which amount of light hydrocarbons
to a solid by cooling. Jet aircraft reduces the opportunities for (light ends) in the product cut.
are frequently exposed to low process optimisation, or they To study cloud point or freez-
operating temperatures and it is need to have an individual ing point, the important var-
essential that their fuels do not analyser for each property they iable is the amount of heavy
freeze in these environments. want to measure (see Table 2), hydrocarbons (heavy ends)
Plugging of filters and related which involves a high capital in the product cut. However,
operational problems in the investment cost in addition to for every model certain stages
fuel system are dependent on high operating costs. However, must be passed to be able to
freezing point; for this reason, it is not common to have so achieve a proper conclusion
jet fuel specifications include many analysers installed in a and results:
requirements for maximum single hydrocracking unit. The 1. Collecting lab historical data
freezing point. optimisation models should 2. Finding the relationships of
Flash point is the lowest tem- aim to: dependent variables
perature at which vapours • Continuously optimise mid- 3. Distillation (D86) analyser
above a volatile, combustible dle distillate production calibration and validation
substance ignite in air when • Minimise middle distillate 4. Defining the on-stream mod-
exposed to flame. The lower product giveaway. el’s dependent variables
the flash point, the easier it is
to ignite the vapour if an igni-
tion source is present. The List of the tests according to the property for diesel and kerosene
higher the flash point, the safer
the material is to handle. This Product stream Property Test Repeatability (r) Reproducibility (R) Notes
means that the temperature Diesel Flash point ASTM-D93 0.0.029×M 0.071×M M = mean
limit to achieve safe storage Cloud point ASTM-D5773 1.3°C 2.5°C result in (°C)
Kerosene Flash point IP-170 1.4°C 3.2°C
is below flash point. A higher Freezing ASTM-D7153 0.6 °C 0.9°C
vapour pressure corresponds to
a lower flash point. Table 2

www.digitalrefining.com/article/1002285 PTQ Q1 2019 3


5. On-stream model validation.
10
5
On-stream distillation analys-
0 ers (OSA D86) are familiar and
−5 available in refineries to meas-
CP, ºC
−10 ure middle distillate distilla-
−15 Lab Model tion. The on-stream distillation
−20
0 10 20 30 40 50 60 70 80 90 analyser is the backbone of all
Time, days our models. However, we have
ensured that calibration of these
Figure 2a TSREC diesel cloud point (lab vs model) analysers has been done accord-
ing to ASTM-D3764. We have
developed the models as a func-
8 tion of the D86 analyser output
6
from the field and were able to
4
2 achieve excellent results. A sin-
0 gle distillation analyser can be
−2 used to give field data for two
−4 different streams (diesel and
−6
−8
kerosene) in any hydrocrack-
Lab CP

−10 ing operation which cycles in


−12 alternative mode between the
−14 two products, which is the case
−14 −12 −10 −8 −6 −4 −2 0 2 4 6 8
with the SSOT unit. A numerical
Model CP
method was used to predict all
of these properties as a function
Figure 2b TSREC diesel experimental (lab) CP vs model calculated cloud point of distillation (see Table 1).

Diesel cloud point


10
In general, as ambient temper-
5
ature decreases, the solubility
0
of n-paraffins also decreases in
−5
CP, ºC

the middle distillate fraction.


−10 Lab Model Precipitation occurs over a wide
−15
0 10 20 30 40 50 60 70 80 90 temperature range until solidi-
Time, days fication finally takes place. The
amount of n-paraffins in mid-
Figure 3a SSOT diesel cloud point (lab vs model) dle distillate has a direct effect
on the fraction’s cold proper-
10 ties (cloud point, pour point,
and cold filter plugging point).
The cold flow property (mainly
5
cloud point) is also important
from an economic point of view.
0
We have found that the diesel
cloud point correlates very well
−5 with the heavy end of the die-
Lab CP

sel distillation (T70% and T95%)


−10 in both hydrocracking units. For
−10 −8 −6 −4 −2 0 2 4 6 8 TSREC, a total of 25 days of sam-
Model CP ple results were used, and in
SSOT a total of 24 days of sam-
Figure 3b SSOT diesel experimental (lab) CP vs model calculated cloud point ple results were used to develop

4 PTQ Q1 2019 www.digitalrefining.com/article/1002285


the correlation model. Sample
collection was done carefully −50
Lab
and precisely to minimise sam- −55
Model
pling error. We were able to do −60
a smart process step change in −65

FzP, ºC
−70
the fractionation section to allow
−75
for variation in both distillation −80
and cloud point and to see the 0 10 20 30 40 50 60 70 80 90 100 110 120
response of the model during Time, days
the days of sample data collec-
tion and models validation. Figure 4a TSREC kero/jet freezing point (lab vs model)
The model’s results for die-
sel cloud point compared to the −50
lab results are shown in Figures
2 and 3 which show alignment −55
between the lab results and the −60
model outcome at any given
−65
time in two different hydroc-
Lab FZP, ºC

racking units. However, Figures −70


2b and 3b show the deviation −75
error between lab measured
cloud point (experimental) and −80
−80 −75 −70 −65 −60 −55 −50
the model CP (calculated). The
Model FZP, ºC
models have an absolute stand-
ard deviation error (TSREC
cloud point, see Equation 1 = Figure 4b TSREC kerosene experimental (lab) freezing point vs model calculat-
1.7°C; SSOT cloud point, see ed freezing point
Equation 2 = 1.4°C). However,
as of today, there is no standard sel in order to save experimen- SSOT kero FzP (°C) = [(T70%) ×
API correlation to estimate the tal time. A total of 38 sample (-0.02213)] + [(T95%) × (0.49808)] -
cloud point from distillation: results were used to develop the (181.68) (4)
model in TSREC and a total of
TSREC Diesel cloud point (°C) = [(T70%) 25 sample results were used in Hydrocracker diesel and
× (0.16427)] + [(T95%) × (0.30471)] - SSOT. Figures 4a and 4b show kerosene (jet) flash point
(165.33) (1) the deviation error between lab Flash point is related to the vol-
measured freezing point (exper- atility of middle distillate and
SSOT Diesel cloud point (°C) = [(T70%)
imental) and the model freez- the presence of light and volatile
× (0.29853)] + [(T95%) × (0.16907)] -
ing point (calculated) for TSREC. components. Equation 5 shows
(160.42) (2)
Figures 5a and 5b show the the API standard method to esti-
Kerosene (jet) freezing point equivalent data for SSOT. The mate the flash point of petro-
We have found that jet freezing models show an absolute stand- leum fractions as a function of
point correlates very well with ard deviation error of 1.21°C for ASTM D86 (T10%):
the heavy end of the jet stream both unit models (TSREC freez-
(T70% and T95%) in both hydro- ing point, see Equation 3; SSOT API Standard Method
cracking units, in exactly the freezing point, see Equation 4). 𝟏𝟏 𝟐𝟐. 𝟖𝟖𝟖𝟖𝟖𝟖𝟖𝟖𝟖𝟖
same way as the diesel cloud As of today, there is no standard !
𝑻𝑻𝒇𝒇𝒇𝒇
&=(
𝐓𝐓𝐓𝐓𝐓𝐓
1

point. We followed the same API correlation to estimate the + [(3.4254) × 10-3 ×
ln(T10) ] - (0.024209) (5)
steps for the jet freezing point freezing point from distillation:
as we had done for the die-
sel cloud point. All of the step TSREC kero FzP (°C) = [(T70%) × where both flash point and T10
changes were done at the same (0.09898)] + [(T95%) × (0.32710)] - are in degrees Kelvin. Equation
time for both kerosene and die- (165.44) (3) 5 predicts flash points with an

www.digitalrefining.com/article/1002285 PTQ Q1 2019 5


to develop a correlation model
−60 to estimate the flash point for
−65 both diesel and kerosene. The
−70 flash point model has an abso-
lute standard deviation error of
FPZ, ºC

−75
Lab
−80 0.82°C for kerosene and 1.7°C
Model
−85 for diesel.
0 10 20 30 40 50 60 70 80 90 100 110 120 Figures 6a and 7a shows the
Time, days flash point model’s results and
Figures 6b and 7b show the lab
Figure 5a SSOT kero/jet freezing point (lab vs model) measured flash point (experi-
mental) and the model predicted
flash point (calculated):
−60

Diesel F.P. (°C) = [(T10%) × (0.5613)] -


−65
(58.868) (6)
−70
Kerosene F.P. (°C) = [(T10%) × (0.6239)]
- (58.448) (7)
Lab FzP, ºC

−75

−80 Kerosene flash point and


freezing point optimisation
−85
−85 −80 −75 −70 −65 −60 −55 A real example of the applica-
Model FzP, ºC tion of the developed model was
the ability to minimise jet prod-
Figure 5b SSOT kerosene experimental (lab) freezing point vs model calculated uct giveaway in hydrocracker
freezing point TSREC. Due to market demand,
jet has higher cracks than the
105 other hydrocracking unit prod-
Lab ucts, whole naphtha and diesel,
100 Model so the main objective is to max-
95 imise the jet production rate.
90 Because the model delivers the
ability to measure both flash
85
point and freezing point contin-
FP, ºC

80 uously, without the time delay


75 required for lab results, we were
0 10 20 30 40 50 60 70 80 90 100 able to increase jet yield by 3%
Time, days (from 42-45%) which is equiv-
alent to 2000 b/d of extra pro-
Figure 6a Hydrocrackers TSREC AND SSOT diesel flash point duction. These tools allowed
operators to see the immedi-
average absolute deviation of However, the estimated flash ate impact on jet product qual-
6.8°C. However, there are some points were always higher than ity with no more waiting until
relations in the open literature our data. So we have devel- the next day to see the results
that correlate flash points to oped our own models to predict of their actions. Now the opera-
either the initial boiling point or flash point as a function of T10% tions team has the ability to see
the distillation temperature at to be in line with API selected full range distillation from the
the 50% point (T50). variables. on-stream distillation analyser,
We have applied the API A total of 30 days of results including flash point and freez-
standard method to both in each hydrocracking unit ing point, at the same time as
streams, diesel and kerosene. (TSREC and SSOT) were used the results are calculated in DCS.

6 PTQ Q1 2019 www.digitalrefining.com/article/1002285


The front line operator can opti-
105
mise main unit fractionator
operation to increase the pro- 100
duction rate by dropping the
95
heavy end of whole naphtha
into jet (for lowest flash point) 90
and pulling the light end of die-

Lab FP, ºC
85
sel into jet (for highest freez-
ing point) without worrying 80
any more about getting off-spec
75
jet. Figure 8 shows the period 75 80 85 90 95 100 105
before and after utilisation of the Model FP, ºC
model. We can now see that we
are able to obtain more data per Figure 6b Hydrocrackers diesel experimental (lab) flash point vs model calcu-
day (four results per day) com- lated flash point
pared to the lab schedule which
was once daily. From day four,
105
the models were applied and Lab
installed in the DCS for a testing 100 Model
period. There was an increased 95
yield of approximately 3% by
90
the end of the testing period,
with the ability to operate close 85
to the highest freezing point
FP, ºC

80
of -47.5°C, compared to -53°C
75
before the model was applied, 0 10 20 30 40 50 60 70 80 90 100
and to be able to minimise the jet Time, days
freezing point giveaway.
Figure 7a Hydrocrackers (TSREC AND SSOT) kerosene flash point
Conclusion
To provide a larger number of
opportunities for optimising the 70
hydrocracking process, mod- 65
els were developed and applied
for the continuous monitoring 60
of physical properties including
55
flash point, freezing point and
cloud point of middle distillates.
Lab FP, ºC

50
The models employed available
45
on-stream distillation analysers
in the field to estimate proper- 40
ties in linear forms. The correla- 40 45 50 55 60 65 70
tion models are very simple and Model FP, ºC
give accurate results about mid-
dle distillate properties with an Figure 7b Hydrocrackers kerosene experimental (lab) flash point vs model
absolute maximum standard calculated flash point
error of less than 2°C out. These
models were derived based on unit. The developed correlations will enable front line operators
distillation data alone, which can will improve both diesel and to monitor target properties and
be easily integrated in DCS with kerosene quality, minimise give- their impact on production rate
the support of the on-stream away of products, and increase continuously. There is no more
analyser in any hydrocracking production rate. The models need to wait for lab results to see

www.digitalrefining.com/article/1002285 PTQ Q1 2019 7


products and lubricants, 2nd ed (ASTM
−46 46 Manual Series: MNL44-2nd), 2007.
−47
45 8 Speight J G, Handbook of Petroleum
−48
Product Analysis, 2002.
−49 44
9 Treese S A, Pujado P R, Jones D S J,
−50 43 Handbook of Petroleum Processing, 2015.
−51
42 10 Hsu C S, Robinson P R, Springer
−52
Handbook of Petroleum Technology, 2017.

Yield, %
−53 FzP 41
11 Wauquier J-P, Petroleum Refining. Vol.
FzP, ºC

−54 FzP model


40 1 Crude Oil. Petroleum Products. Process
−55 Yield
Flowsheets, 1995.
−56 39
0 48 96 144 192 240 288 336 12 Scherzer J, Gruia A J, Hydrocracking
Time, hours Science and Technology, 1996.

Figure 8 TSREC jet data (freezing point vs yield)

Madi Asiri is an Operations


the immediate impact of process 2 Watkins B, Lansdown M, Understanding
Cloud Point and Hydrotreating Superintendent with Saudi Aramco
changes on final product quality. Total Refining and Petrochemical
Relationships Revised, Chicago, IL, USA,
The kerosene production rate 2012. Company (SATORP), Saudi Arabia. He
increased by 2000 b/d in a single 3 Refinery improves diesel cloud point graduated as a chemical engineer from
unit. We believe the developed by switching to Topsoe’s catalysts, Haldor King Fahd University of Petroleum &
models can be applied in any Topsoe, cvr 41853816, CCM, 0063.2015/ Minerals (KFUPM), Saudi Arabia, and
hydrocracking unit and will give Rev.0, 2015. holds a graduate diploma in refining and
excellent prediction of the tar- 4 Hansen J A, Andersson A S, Optimising petrochemicals from IFP School, France.
get properties cloud point, flash middle distillate production in a Email: madi.asiri@satorp.com
point and freezing point utilis- hydrocracker, 2017.
ing a single on-stream analyser 5 Dinkov R, Stratiev D, Penev D, Cholakov LINKS
for both diesel and kerosene. G, Investigation On Diesel Cold Flow
Properties, Bulgaria, 2010.
More articles from the following
Further reading 6 Riazi M R, Characterization and
categories:
1 Zhu F (Xin X), Hoehn R, Thakkar V, properties of petroleum fractions, 1st ed
Coking
Yuh E, Hydroprocessing for Clean Energy: (ASTM Manual Series: MNL50), 2005.
Instrumentation, Automation
Design, Operation, and Optimization, 7 Kishore Nadkarni R A, Guide to ASTM
and Process Control
2017. test methods for the analysis of petroleum

8 PTQ Q1 2019 www.digitalrefining.com/article/1002285

Вам также может понравиться