Ethylene and Propylene are two of the most important petrochemical intermediates. They are the building blocks for producing major polymers such as polyethylene, polypropylene and various other petrochemical products.

Steam Cracking of naphtha or LPG  is the most common way of producing these chemicals. Typical Ethylene plant has several furnaces (aka Hot section), where the feedstock mixed with steam is cracked at temperatures ranging from 820 to 860 degC. The cracked gas is then separated in series of columns, (aka “Cold section”).

There are several different configurations of ethylene plant currently in operation.  Following is the block diagram of an typical ethylene plant (with deethanizer as a first column in the cold section).

The cracking severity is measured by P/E (Propylene/Ethylene ratio) in the cracked gas. The feed composition is one of the main disturbances that affects the severity. The cracking kinetics is well understood, and several first principle based models are available to model and predict the cracked gas composition. These models take the feed composition and furnace process conditions as inputs and calculate the cracked gas composition.

However, there are some issues to use the predictions by first principle based models directly for control:  

1.      First principle based models do not provide the linearized gains between the manipulated variables and controlled variables

2.      First principle based models usually don't come with data –preprocessing. (Filtering of inputs/outputs, validation of inputs) 

One way of tackling these issues is to build regression models, from the data generated from the offline simulation of first principle based models. The regression models are more robust against measurement errors, and linearized gains can be extracted from the models. The regression models are updated by the first principle models.

Then, APC (Advanced Process Control) takes the information given in the regression model -- the prediction -- and uses it as a controlled variable. (To learn more about this, read previous articles on the benefits of APC here -- and APC fundamentals here)

The following illustration shows the typical input and output of different modules and how they are integrated with each other. (RQE – Robust Quality Estimator, SMOC – Shell Multivariable Optimizing Controller are packages of the Yokogawa APC Suite).

Plant Wide Control and Optimization.

Plant wide control involves linking several sub models of the process, under a single main controller/optimizer. The intermediate variable that connects the sub controllers is typically the feed flow from the upstream sub controller. The overall main controller optimizes the whole plant taking into account of constraints across the whole plant.  Following are some of the examples of how plantwide control concept can be applied to maximize the benefits.

Are you enjoying this article?? Follow us on LinkedIn and Facebook.. 

Furnace Optimization:

The cracking reaction result in “coking” of furnace coils, which affect the heat transfer. Hence the required heat-duty to operate the furnace at the desired COT (Coil Outlet Temperature) increases with time. The TMT (Tube Metal temperatures) increases gradually reaching the metallurgical constraints. The furnace is then taken out of service for “decoking” after certain period of operation, typically around 2 months.

When a furnace is near “End of Run,“ the constraints on TMT must be carefully monitored, and furnace severity or load need be adjusted so as not to violate the constraint on TMT.  APC using “Plant Wide Control Concept” can manage this efficiently and the overall main controller can optimize the load balancing or yield across the furnaces subject to TMT constraints in different furnaces.

The above concept can also be applied when “energy savings” is a concern. The load balancing is then based on the fuel efficiency of the furnace. 

Cold Section Optimization:

There are several constraints that are encountered in the cold section, which are managed by manipulating the process conditions in the hot section. For example, when the LPG is cracked, cracked gas contains higher proportions of lighter components than when naphtha is cracked. In this case, some downstream columns could be constrained in terms of reboiler duty.

The plant wide controller can manage such constraints more effectively. For example, the overall main controller adjusts the feed rate to individual furnace, feed mix (relative proportion of naphtha and LPG to be cracked) when the deethanizer reboiler steam valve is saturated.

Another example is the minimization of steam consumption in the refrigeration compressors. Typically the feed flow is taken as a “Disturbance variable” for a column. However, in the plant–wide controller the intermediate product flows are predicted by the main controller and passed to the sub controller for dynamic optimization. This avoids the “sub optimal” control moves. For example, the reflux is minimized in advance, based on the prediction of feed flow from the main controller, which results in less energy consumption – such as reduction in refrigerant duty or reboiler duty. 

Case Study:

In 2012, Yokogawa implemented SMOC (Yokogawa/Shell’s APC Suite) in an Ethylene plant for a leading petrochemical producer in Thailand. The plant cracks Naphtha, LPG and recycled ethane. The plant is designed to produce 300 KTA of Ethylene.

One main controller with 16 sub controllers was built to meet the control and economic objectives. Cracking Kinetic model SPYRO (from Technip) was used to predict the cracked gas composition. The overall cracked gas effluent composition (ethylene and propylene content) was calculated from the individual furnace effluent predictions. The sub controllers were linked by “feed flow” intermediate variables.

The main controller maximized “Ethylene” yield subject to the constraints across the whole plant. The post-implementation audit showed a benefit of more than US$ 3 million per annum.

The main benefit area can be seen in the chart below. Also take note that in addition to the economic benefits, the operator acceptance of the APC was quite high, resulting in uptime of more than 90%. 

Do you have a question about this article? Please CONTACT US.