MODERN FUEL BLENDING-2 DIESEL BLENDING SYSTEM CUTS QUALITY GIVEAWAYS AT RUHR REFINERY

Yalcin Serpemen, Fritz W. Wenzel
Veba Oel Technologie GmbH
Gelsenkirchen, Germany

Alfred Hubel
Veba Oel AG
Gelsenkirchen, Germany

An advanced diesel/gas oil blending system technology employed at Ruhr Oel GmbH's Gelsenkirchen-Horst refinery has improved blending operations and reduced costs in several important areas.

In addition to a gasoline blending system (OGJ, Mar. 18, p. 62), Ruhr Oel has operated an on-line blending system for gas oil (diesel and light heating fuel oil) since December 1989 at its Gelsenkirchen-Horst refinery.

Changing diesel fuel specifications in the U.S. and Europe will put more emphasis on such systems.

As stated in the first article, an on-line blending system fulfills the most important blend objectives by:

Minimizing quality giveaways and reblends
Producing the most valuable products at lowest possible costs by minimizing the usage of expensive high-value components, such as reformate or purchased components (oxygenates), in the case of gasoline; or desulfurized gas oil and additives (flow improvers), in the case of gas oil
Providing a system with accurate quality control using on-line analyzers
Increasing the blending operation flexibility
Reducing the inventory of components and finished products.

In the case of diesel and light heating fuel oil, the future reduction of aromatics and sulfur content impacts the chemical composition of the pool components. The most likely effects will be on cold filter plugging point (CFPP) and cloud point additives.

BLEND SYSTEM

The main features of the blending system are highlighted below using the Ruhr Oel Gelsenkirchen installation for gas oil blending as an example.

Because of the complex structure of this petroleum refinery (Fig. 1, OGJ, Mar. 18, p. 62), with integrated petrochemicals production, the gas oil pool consists of components with very different properties.

These components are blended into two different grades of gas oil (diesel and light heating fuel oil), with seasonal quality changes. Up to nine single components, as listed in Table 1, plus cloud point and flow improvers, are involved in each blend.

The system configuration, as schematically shown in Fig. 1, consists of:

omponent tanks (once-through, bypass, and pre-blend modes)
Additive tanks
Transfer lines with flowmeters and control valves
One blend header with a capacity of approximately 80,000 b/d, with a special integral in-line mixer providing a homogeneous blend of all involved components
A process computer (which also handles the gasoline blending system) as a central supervising unit, collecting all data (including product tank heel volume and quality), calculating the optimal recipe, and downloading it directly to the ratio controller
A ratio controller (in-line blender) controlling and transmitting the volumetric throughputs of the components during the blend.

Because of the special feature of operating with nonstatic component tanks, the system has to perform simultaneous quality and quantity integration of the components. This configuration of the blending unit allows for blend control in a closed-loop mode.

The blending optimization and supervisory system applied by the process computer uses on-line and real time multivariable LP-based optimization techniques.

The system communicates by means of interfaces with the ratio controller, the online analyzers at the blend header and process units, and the automatic tank gauging (ATG) system.

It provides actual information on components and finished product tanks during the blend.

The blend operation and control is performed by means of the "operators interface," providing the plant's shift personnel with all necessary functions (blend setup, start, stop, suspend, etc.) and information (displays and status and trend reports), including specific alarms during the actual blending operation.

The setup of the control room is illustrated in Fig. 2. The functions and elements of the blend optimization system and control system are shown in Fig. 3.

During the blending operation, the on-line analyzer measurements at the fast sample loop are used, after the completion of a validity check, to update the blending model periodically. A quality integration of the blended product (in the product tank) is performed using the online analyzer readings and taking into account heel volume and qualities.

Within the given constraints of, for example, the equipment, component availability, component quality, and final product qualities, the LP-optimizer ensures the most economic usage of the given components. In case of infeasible operating conditions in the plant, the operator is guided toward a feasible solution.

The result of the calculation is then directly downloaded to the component ratio controller.

Because the system has to operate with nonstatic tanks, the knowledge of changing component qualities is important.

This is especially true because the refinery tends to charge different crude oils frequently (every 2 days, on the average). Therefore, already existing analyzer readings at the process units are used to update the component qualities (Fig. 4).

The integrated quality of the blended product in the product tank, including the volume and quality of the tank heel, is calculated every 6 min using the on-line measurements. This results in new quality set points, calculated by the system, for achieving the required values in the tank.

Because the on-line analyzers in question need different lengths of time to generate data points, synchronization of the analyzer readings has to be performed. Therefore, the instrument with the longest response time dictates the cycle time for LP-model update and the setting of new ratio values at the blend header.

On-line analyzers for measuring the density at 15 C. (D15), cloud point (CLP), sulfur content (S), flash point (FLP), cetane index (CI), and cold filter plugging point (CFPP) of the product at the blend header are used for this application.

This allows for the feedback control of D15, CLP, S, and FLP. Additive insertion is not measured by means of an analyzer, but controlled by volume-proportional addition to the blend.

The on-line process analyzers embedded in a properly designed fast sample loop are essential for achieving a high blend optimization performance. Therefore, two main requirements have to be fulfilled during the blending operation:

The on-stream factor, or the availability of the analyzer must be high, providing blend control capability by means of blend quality corrections.
Accuracy must be high at the same time, allowing for a blend quality control as close as possible toward the targets, reducing quality giveaways.

A fast sample loop design, accounting for proper sample conditioning for the integrated on-line analyzers, along with calibration facilities, can contribute significantly to achieving the previously mentioned requirements for the analyzers.

RESULTS

The related cost savings resulting from the reduction of cloud point and sulfur giveaway, the maximization of the density spread between diesel and light heating fuel oil (because of the German taxation system), and the reduction in additives and flow improvers can easily be calculated.

These cost savings have been the basis for justifying the gas oil blending project for Ruhr Oel's Gelsenkirchen refinery.

However, additional considerations can also contribute appreciably to the economic improvements. These are credit potentials resulting from the elimination of reblends, optimal tuning of the cut point between diesel and atmospheric gas oil, and the possible reduction of laboratory work (certification).

Quality give-away, the extent of required reblend operations, and laboratory effort are strongly interrelated. Quality give-away usually results from the limitations on time available for performing reblends for correcting either off-spec qualities or quality giveaways.

Fig. 5 shows a broad view of the gas oil blending system at the Gelsenkirchen-Horst refinery.

Fig. 6 gives an impression of the inner part of the analyzer house with the CFPP, CLP, and FLP analyzers on the left-hand side, and the density and sulfur analyzers on the right-hand side.

START-UP PROBLEMS

The following start-up problems occurred and influenced the overall performance of the system in the first months.

During winter operation, flocculation of additives was observed, resulting in some problems in the metering system and some plugged filters.

Increasing the tracing temperature solved this problem.

The temperature control of the fast sample loop had to be optimized to avoid temperature fluctuations in the supply streams to the analyzers.

Due to unscheduled shutdown of the Molex unit for approximately 6 months, the pool composition had to be drastically changed, as compared to usual operation. Therefore, some recalibration of the on-line analyzers was necessary during start-up.

The impact of the above problems in the 1990 overall performance has not been corrected in the data shown. Therefore, improved performance in the coming years is to be expected.

ON-LINE ANALYZER PERFORMANCE

As a result of the 1990 operation, the following on stream factors (volume-based average values) have been achieved: gravity, 99%; cloud point, 97%; sulfur, 99%; flash point, 99%; and CFPP, 98%.

As an example, the availability for the CFPP analyzer plotted over the number of batches is given in Fig. 7. This high on stream factor provides a good basis for a successful blend quality control.

As mentioned before, the second important factor for successful quality control is related to analyzer accuracy, in conjunction with volume-based tank quality integration.

Because both the laboratory method and the on-line analyzers are measuring to standards (ASTM and DIN, respectively), the deviations of the measured values have to be within the defined comparabilities.

The second column in Table 2 gives the comparability limits (allowable deviations) of the applied standards. Column 3 gives a comparison of the system prediction with the laboratory results.

The system prediction is the result of on-line analyzer readings integrated over the tank volume, including consideration of the tank heel. The on-line results are, depending on the value measured, between 87 and 100% within the limits of comparability with the laboratory's results.

Fig. 8 shows the deviation of cloud point measurements made by the laboratory with those predicted by the system. As can be seen, the deviation is within 1 C., which is far more precise than required by standards.

With respect to sulfur measurement, approximately two thirds of all values (laboratory vs. system prediction) are within 0.01% accuracy. Frequent recalibration significantly improved the accuracy, increasing it up to this level.

Sample conditioning has been the key issue for the flash point analyzer. Today's operation, due to improvement in this area, shows very satisfying results.

The CFPP measurement is not actually used within the optimization routine. This is because the available calculation routine to predict the interrelationship between additive addition rate and effect on CFPP is not yet satisfactory. Therefore, the analyzer reading is used to manually adapt the rate of the flow improver to the blend header. In addition, the readings are used for quality integration calculations with acceptable results.

Even so, a 50% reduction in flow improver has been achieved, as compared to the former operation. Nevertheless, this is an area where further improvements could be expected.

The cetane index is calculated by using the measured gravity and the boiling curve of the different blend components (from the system's data bank). Reasonably good results have been obtained; within the comparability limits of the cetane index and cetane number (laboratory measurement).

In summary, the good results from the on-line analyzers have already resulted in reduced laboratory work, which is expected to be further decreased.

REDUCTION OF QUALITY GIVE-AWAY

The on-line blend optimization system described here is able to control all qualities simultaneously by using online real time multivariable LP-based optimization techniques. This results in a new recipe every 6 min, producing a significant reduction in quality give-away.

The following values are averages for the first year of operation: cloud point deviation, 1 C.; sulfur content deviation, 0.02 wt %; gravity spread, 4 kg/cu m.

REDUCTION IN OPERATING COST

With respect to additives in general, a reduction of 50% has been realized. Furthermore, almost no reblend has been necessary, resulting in lower operating cost (energy) and higher utilization of product tank volume, as no allowance for spare volume for reblends is necessary. Subsequently, laboratory work has been reduced approximately 65%.

A more efficient utilization of blend components (as reported for gasoline blending) has not been realized at present. This is because two major low-paraffin pool components (Molex raffinate and the C13 stream) were not available for more than 6 months in 1990.

OUTLOOK

As shown in Fig. 4, several on-line measurements are available from different process units to improve the knowledge of component qualities.

In 1991 these values will be fed directly to the system for continuous update of the component qualities. This will allow a reduction in safety margins on the product side, which will result in an additional decrease in quality give-away.

In view of the new gas oil sulfur specifications in Europe from 1993 onward (0.05%, as compared to 0.2% at present), the accuracy of the sulfur measurement has to be further improved.

The establishment of a correlation between CFPP value and additive addition should result in a further decrease in the use of flow improver. Upon the full availability of all normally available pool components, an optimized component utilization is expected.