|
|
| Ineos Phenol, Gladbeck (Germany) | | Advanced Process Control at Ineos Phenol | Ineos Phenol world's biggest producer of Phenol
Since 1954 Phenol and Acetone are produced in the site Gladbeck, Germany. At that time the production started with 8000 t/a Phenol and 5000 t/a Acetone. Meanwhile the production increased because of expansions and improvements to 650,000 t/a Phenol and 400,000 t/a Acetone. Therewith the site in Gladbeck, Germany is the biggest production site worldwide for Phenol. With the additional sites in Antwerp (Belgium) and Mobile (USA) Ineos Phenol is the world biggest producer of Phenol.
The process of Phenol production is at Ineos Phenol continuously improved to increase the plant efficiency, decrease the consumption of raw material and energy as well as the air and water emissions and to develop the general plant safety to the highest industrial standards. A part of the improvements is the implementation of a model predictive controller at the Gladbeck site.The production process of Phenol consists of two main steps. In the so-called Oxidation the raw material Cumene is converted via oxidation and cleavage to Phenol and Acetone. In the following distillation the main products Phenol and Acetone are separated. The separation occurs in a complex composite of rectification columns thereby also the side products of the reaction are segregated. This composite of columns is characterized by a level of high heat integration, a strong coupling of mass flows and a major number of recycles. The operations of some central columns affect the efficiency of the whole distillation and the quality of the produced products.The behavior of one of these central columns is characterized by a strong coupling between manipulated and controlled variables. Due to these interactions and the large time constants in which the column reacts on changes of the manipulated variables, some deviations could only be compensated in form of manual interventions in the past. With the implementation of the improved process control at this central column Ineos Phenol was overcoming this situation and increased the automation level. Two Projects for efficiency increaseThe improvement of process control was carried out in two sub-projects. In a first step Siemens was commissioned to carry out a PID tuning based on the existing control loops at several points in the plant. Consequently the Siemens specialists analyzed control loops in the distillation part and further more worked out adequate optimized control concepts. For this purpose Siemens used a tool to detect optimized control parameters. After that the new control parameters were implemented in the existing control system. At the same time a control concept based on a model predictive controller was developed to optimize the operation of one of the central columns.Parallel to the PID controller tuning Ineos Phenol came to the decision to migrate hot plug-in the existing control technology to the process control system Simatic PCS 7. Due to the open system structure of PCS 7 it was possible to transfer the results of the controller tuning trouble-free as new controller parameters to PCS 7. Due to good results of the controller tuning Ineos Phenol decided to give the realization of the automation concept for one of the central columns based on a model predictive controller worked out in this step into the hands of the Siemens specialists. How a Model Predictive Controller worksA model predictive controller (MPC) is a control concept that belongs to the group of advanced process control (APC). Such control concepts are normally based on the existing basic automation with PID-controllers. With APC control concepts the control of the plant is enlarged, existing simple control loops persists. A model predictive controller is based on the schema of a master and slave controller. The model predictive controller is the master controller that forces set points for several subordinated PID-controllers, the slave controllers, depending on different controlled variables. Thereby the fast automation functions like e.g. holding and adjusting the throughput to a set point by manipulating valve position are performed by the therefor designed basic automation. The advanced and complex control tasks are executed by the model predictive controller. Typically a model predictive controller works with a time constant in minute range (PID-controller typically in millisecond range).A model predictive controller is a model based controller. In contrary to PID control loop (single variable control with one manipulated and one controlled variable) in a model predictive controller several manipulated and controlled variables are evaluated at simultaneous The time dependent reaction of all controlled variables to changes of the different manipulated variables is mapped in the process model. With this dynamic model it is possible to predict the future behavior. By means of a mathematic optimization procedure and the model the optimized trajectory of the manipulated variables are calculated in the model predictive controller in a way that the defined control targets are reached as fast and comprehensive as possible. This optimization is repeated at every cycle. The first part of the optimized manipulated variable trajectories are implemented as set points to the subordinated PID-controllers.The control targets of a model predictive controller are not only set points for the controlled variables. Typically control corridors (target ranges) are defined for the controlled variables. If the controlled variable is located inside the corridor, the control target is reached and no manipulated variables have to be changed. Additionally it is possible to specify targets for the desired position of the manipulated variables if the corridors for the controlled variables have been reached. Therewith it is possible to integrate advanced issues such as economic targets into the control concept. With the model predictive controller used by Siemens the control targets can be classified hierarchically. First priority is to bring all controlled variables into the target corridor. If these targets are reached, it can be achieved to fulfill further economic targets like the reduction of the energy consumption (e.g. steam), without violating the before reached targets. All control targets can be combined in any way and sorted into as much as needed hierarchically levels. Integration of MPC into Simatic PCS 7The operation of the model predictive controller is separated into two levels. At first the model predictive controller has to be configured. This engineering is done with the tools of the controller software and is carried out at Siemens by APC specialist. In a second step the model predictive controller is running in normal operation by the operators. This handling is carried out in Simatic PCS 7 by using the normal operator screens. To connect the model predictive controller with Simatic PCS 7 there are so called link modules available. These link modules have only to be configured (like the parameterization of a PID block). Further programming is not necessary. Beside others these link modules allow the supervision of the operativeness of the model predictive controller (watchdog) and the combined switch-on of all APC variables. Further more for every manipulated, controlled and disturbance variable block icons and faceplates exist, at which necessary operator interaction for the model predictive controller are carried out (set points, set point corridors, use/not use variable, ...). The operator can carry out every operation in a familiar environment. No additional operator stations are necessary. Model Predictive Controller at Ineos PhenolBecause of the simultaneous replacement of the control system by the new DCS Simatic PCS 7 the implementation of the model predictive controller was carried out in two steps. In the first part the plant behavior was identified by plant tests. This work was done with the old control system. With the results the model for the model predictive controller was developed. The model describes the influence of four manipulated and two disturbance variables on two controlled variables. The implementation of the model predictive controller during plant operation was done in the second step after the implantation of Simatic PCS 7.By means of the model predictive controller an important distillate concentration in the central column is held constant. At the same time the bottom temperature is kept in the specified temperature range. Beside the reflux and the steam flow to the bottom of the column an additional feed to the column is used as manipulated variables. Additionally the heat integration with a second column is incorporated. The controller has the ability to detect early trends and to react adequately. After a project duration of 5 months the APC solution was approved by Ineos Phenol in January 2006. An advantage of the work with the new controller is the seamless integration into the operation of Simatic PCS 7. With only one mouse click the operator can open the relevant mask. Because of special link modules, the model predictive controller could be easily integrated into PCS 7 without any additional programming effort. With these link modules the controller depended values and parameters and also the functional reliability of the communication connection with the model predictive controller can be observed. Thereby at the event of a malfunction a previously defined fall back strategy can be activated and so expensive failures of the controller can be avoided.This new control solution has enabled Ineos Phenol to successfully stabilize the critical distillate concentration. Ineos Phenol is very satisfied with what has been achieved and appreciates the excellent collaboration with Siemens.
published in: Hydrocarbon Engineering 10/06 |
|
|
|
