Fuzzy Logic Supervisory Control for Coal Power Plant

by Dirk Pruessmann (3/1997)

Citation Reference: The Fuzzy Logic Application Note series is published by Inform Software Corporation on its Internet server to promote the use of fuzzy logic technologies in applications.

The High Temperature Winkler Gasification (HTW) process that was developed by Rheinbraun AG has been used for many years in pilot and demonstration plants to generate synthesis gas and fuel gas out of brown coal. Conventional methods were used before to control the gas throughput. While the conventional control engineering implementation was able to run the process in a stable operating point, improvements were necessary to use the HTW process in a coal power station with integrated coal gasification:

On top on the existing base level automation, a supervisory fuzzy logic control strategy was implemented on the HTW plant in Berrenrath/Germany. Fuzzy Logic was used because the control problem was strongly non-linear and involves multiple measured and command variables. On the other hand, extensive operator knowledge about the process was available. The implemented fuzzy logic supervisory control strategy successfully improved throughput control quality as well as the adaptation to different coal parameters.

1. High Temperature Winkler Gasification

The process that is used to gasify the coal is called High Temperature Winkler method (HTW). The HTW Gasification method uses a high temperature fluid bed process to convert brown coal into synthesis gas, a mixture of carbon monoxide (CO) and hydrogen (H 2 ). This gas mix can be used to produce chemical base products like aldehydes or organic acids. Alternatively it can be used in a power plant gas turbine to generate electricity. The gas produced by the demonstration plant is used for chemical synthesizes. Later in the power plant application the gas will be used to run a gas turbine / steam turbine combination.

The HTW process has been used for the gasification of coal by the German coal company Rheinbraun since 1956. The demonstration plant started operation in 1985. It converts 720 t of coal per day into 900.000 m 3 (iN) synthesis gas. In 1996 Inform added a fuzzy logic supervisory control to enable the process for a power plant application. Figure 1 shows a photo of this plant.

HTW Plant Figure 1: HTW Plant in Berrenrath, Germany

The main inputs for the HTW process are coal, oxygen and steam. The coal is first ground to small pieces and pre-dried before it is fed into the bottom part of the fluid bed reactor. The steam and the oxygen are fed into the reactor on four different levels, into and above the fluid bed. In the fluid bed the coal reacts with the oxygen and the steam. This reaction takes place at a temperature of around 800°C and at a pressure of 10 bar. After the reaction in the fluid bed the generated gas enters the hot zone above the fluid bed. At temperatures around 1000 °C additional oxygen and steam is added and left over coal particles react with the gases. This way additional gas is produced and by-products like methane and other hydrocarbons are converted to carbon monoxide and hydrogen. The produced gases leave the reactor at the top through the reactor head. At this point the gas is still mixed with a lot of particles. These are filtered out and fed back into the fluid bed with a zyklon filter and a feed back tube.

Process Diagram Figure 2: Process Diagram of HTW Plant


The coal ashes accumulate at the bottom of the fluid bed reactor. They are removed from there out of the reactor by two conveyor spirals. The hot raw gas is cooled down to 270°C. Its heat is used to generate pressure steam, some of which is recycled back into the process. Ceramic filters remove remaining dust particles out of the gas. The following gas washer removes NH 3 , HCL and other gas components. The following CO conversion creates the correct carbon monoxide / hydrogen mix for the methanol synthesis. After a compression to 37 bar the gas is processed in a non-selective rectisol washer (C0 2 /H 2 S washer). At temperatures below -40°C liquid methanol is used to wash out carbondioxide and sulfuric components. The methanol is used again after recycling it and the purified synthesis gas is used at a nearby chemical plant. Figure 2 shows the process diagram of the HTW plant.

The coal input, the oxygen input, the distribution of the oxygen input over the eight different nozzles and the ash removal rate have to be controlled to use the coal efficiently and to generate the correct mixture of gases.

Instead of coal a mixture of coal and plastic refuse can be used in the HTW process. This way the coal consumption is reduced and the plastic refuse is recycled into synthesis gas.

2. Conventional Control

The HTW demonstration plant is controlled with an Eckhardt PLS-80E DCS system. This system controls over 6000 measurements and actuators. The main control room is equipped with 10 Unix based operator consoles and four real-time servers. So far nearly all the set points of the underlying control circuits are set and adjusted manually by the operators. They constantly monitor the process condition and adjust the set points of the underlying control circuits accordingly (i.e. coal input, oxygen input, ... ) A few years ago it was tried to automatically generate some set points using a conventional PID controller But the results were not satisfying. This supervisory control only worked fine when the coal quality was very constant. Otherwise the process quality would deteriorate significantly and the operators had to intervene and switch back to manual operation.

3. Supervisory Fuzzy Logic Control

Two main tasks have been defined for the supervisory fuzzy logic control: Regulation of the gas throughput and process stabilization.

The fuzzy logic must keep the gas throughput at the set point and it has to respond to set point changes with the correct dynamic speed. The set point can vary from 70% (partial load) to 100% (full load). The fluctuations of the gas throughput result mainly from variations of the coal quality (humidity, ash content and granularity). These effects have to be compensated by the fuzzy logic.

The process stabilization must keep several process parameters in the optimum range. The reactor load influences the optimum of these process parameters. The position of the optimum also depends on the coal quality. The following parameters were used to define the quality of the process:

The process stabilization is especially difficult when the HTW process is fed with a mixture of coal and plastic refuse. This is done because plastic refuse is a very inexpensive fuel. But the addition of plastic to the coal results in drastically different process conditions. The fuzzy logic control uses the regular measurements of the process conditions to detect any addition of plastic to the coal. This will result into an adapted control strategy of the fuzzy logic control.

4. Fuzzy Logic Control Design

The specification of the control task resulted into a preliminary concept of the fuzzy logic controller. The operator knowledge was than used to specify the control strategy of the fuzzy logic system. Several structured audits took place to evaluate the operator knowledge systematically. The audits focused on the operators' manual control strategies and on the relationships between the inputs and the outputs of the process. This procedure is in accordance with the standardized fuzzy logic design method.

The audits resulted into the following concept for the fuzzy logic control: Deviations of the gas throughput from its set point immediately result into a correction of the oxygen input. The coal input is adjusted accordingly to keep the ratio between coal and oxygen at a constant level. Changes of the reactor pressure predict changes of gas throughput. Therefore the pressure gradient is used as an early warning indicator for changes of the gas throughput.

For process stabilization and for the adaptation to different coal qualities the fuzzy logic controller uses the following parameters to keep the process in a stable operating condition:

Sometimes one and the same process input value has to be modified to keep several different critical measurements (temperature, fluid bed height, ...) in the optimum range. This can have conflicting results.

For example: A too low fluid bed is normally corrected with an increase of the coal input. A too low temperature is corrected with a reduction of the coal input and a too low dust output is also corrected with a reduction of the coal input. But a too low temperature can occur together with a too low fluid bed. The ability of the fuzzy controller to weight different conflicting indications based on their significance and to use a lot of inputs to determine the best reaction to each situation is very useful to control complex processes.

The fuzzy logic controller has a total number of 24 inputs and 8 outputs. A preprocessing reduces the 24 inputs to 10 characteristic descriptors. These are fed into the fuzzy logic system. The fuzzy outputs go through a post-processing step to generate the actual set points for the process inputs. Figure 3 shows the core structure of the fuzzy logic system.

Fuzzy Controller Structure Figure 3: Fuzzy Controller Structure

5. Integration of Fuzzy Logic into the DCS

The process measurements are coming to the fuzzy logic control through the Eckhardt DCS. The fuzzy logic system generates set point values for the underlying PID controllers. The fuzzy logic controller was implemented on an OS/2 PC. Therefore the set up of the communication between the DCS and the fuzzy logic controller was an important part of the whole project. The communication was implemented using Factory Link, a well-known SCADA program by US-Data. A factory link application is running on the same OS/2 PC together with the fuzzy logic controller. The fuzzy logic controller reads data out of the Factory Link Real Time DataBase and it also writes data back into it. Another Factory Link task is communicating with the DCS through the Eckhardt DCS bus. This way an image of the process measurements is created in the Factory Link RTDB and the fuzzy outputs are forwarded to the DCS.

Factory Link initiates a new fuzzy logic evaluation every 10 seconds. The DCS either uses the external set points generated by fuzzy logic or the internal set points entered by the operators. The operators can switch from the 'manual mode' to the 'fuzzy logic mode' and back. The 'fuzzy logic mode' can only be activated when all the critical system variables are in a predefined safe range. The DCS automatically switches back into 'manual mode' whenever a system variable exceeds the safe range. The fallback to 'manual mode' also takes place if the communication between the DCS and Factory Link is interrupted.  

Integration of Fuzzy Logic Figure 4: Integration of Fuzzy Logic into DCS


Figure 4 shows the integration of the fuzzy logic control into Factory Link and the Eckhardt DCS.

The OS/2 PC is also connected with a serial cable to a WIN95 PC, on which the fuzzy TECH development system is installed. This program was used to develop the fuzzy logic control and to generate C-Code for the implementation on the OS/2 PC. The WIN95 PC is also used for On-line optimization and visualization of the fuzzy logic controller. The serial link to the OS/2 PC enables the user to modify the fuzzy system on the fly from the fuzzy TECH development system on the WIN95 PC while the system is running and controlling the process.

6. Setting the Fuzzy Logic Control into Operation

The first design of the fuzzy logic control, the data preprocessing and post-processing were tested using the simulation tool VisSim. This way the concept was checked for any structural errors and an early prototype was presented to the customers. Figure 5 shows a test of the fuzzy logic system using simulated data as input.

Simulation with VisSim Figure 5: Simulation with VisSim


After the successful completion of the simulations the fuzzy logic was tested off-line with real time data from the DCS. To do this the fuzzy logic control first was implemented on the OS/2 PC. The fuzzy controller than used real time DCS measurement values to generate set points values for the DCS. But during these off-line simulations the DCS was only using the internal manual set points and not the fuzzy logic set points. By comparing the external fuzzy logic set points with the internal operator set points deviations between manual and automatic operation could be detected and if necessary eliminated.

After the off-line testing was finished successfully the on-line testing started. For the on-line tests the DCS activated the external set points and so the closed loop performance of the fuzzy logic controller could be tested. The fuzzy logic controller was optimized while running in the closed loop mode from the fuzzy TECH development tool on the WIN 95 PC. To do this the OS/2 PC was connected with a serial cable to the WIN 95 PC. This way the fuzzyfication, inference and defuzzification were visualized in fuzzy TECH. Modifications of the rule base or term definitions were also entered in fuzzy TECH and than send to the fuzzy logic controller on the OS/2 PC.

7. Results

The fuzzy controller proved to be working very effectively during the first few on-line tests. After that a long series of evaluation tests started. During these tests the performance of the fuzzy controller was tested using a lot of different coal qualities and different loads (70% to 100%).

Regulation of gas throughput

The conventional control focussed on process stabilization not keeping the throughput at its set point. The following diagram shows the changes in the gas throughput when fuzzy controller is active. The fuzzy logic control clearly improves the throughput control. It compensates fluctuations in the granulation of the coal.

Figure 6 shows the fluctuations of the synthesis gas throughput with and without fuzzy logic.

Throughput Control> Figure 6: Throughput Control with and without Fuzzy Control


Change of Load

The fuzzy controller has to regulate the gas throughput in the range 70 % to 100 % load with a maximum load gradient of 4% / min.

Load with Fuzzy Control Figure 7: Change of Load with Fuzzy Control

Figure 7 shows a load change from 94 % to 76 % and back to 94 %. The resulting load gradient was 3,2 % /min. Currently the load change behavior is being improved by using a modified pressure evaluation.

Adaptation to coal add-ons

Sometimes the HTW process is fed with a mixture of coal and plastic refuse. The resulting process parameters vary greatly from the standard operating conditions. For example the content of methane in the raw gas increases. The fuzzy controller recognizes the different operating conditions and generates a matching internal set point for methane. Figure 8 illustrates how this enables the fuzzy controller to stabilize the process.

Process Stabilization Figure 8: Process Stabilization

Adaptation to different coal qualities

The fuzzy controller also keeps the synthesis gas throughput constant when the coal quality changes. Figure 9 shows the results of a change of the coal's water content from 18 % to 12 %. The fuzzy controller reduces the coal input to compensate the coal change. Later after switching back to moist coal the coal input is increased accordingly.

Throughput Figure 9: Constant Gas Throughput with Different Coal Qualities

Conclusion

The fuzzy controller was implemented very quickly. The transfer of the process know-how into the fuzzy controller and its realization took 15 days. So far over 1100 hours of operating time have been evaluated. In nearly all situations the performance of the fuzzy controller was much superior to the manual control. It was able to keep the process parameters in the optimum range whenever the coal quality changed. It was also able to adjust the gas throughput with the necessary change rate. The average gas throughput was kept at the set point. The operating personal has accepted the fuzzy controller as a helpful component because its transparent integration into the PLS makes it easy for them to use it.

8. Literature

[1] Evans, G.W.; Karowski, W.;Wilhelm, M.R., "Application of fuzzy set methodologies in industrial engineering", Amsterdam, Oxford, New York, Tokyo (1989).
[2] N.N., " fuzzy TECH 4.2 Manual", INFORM GmbH Aachen / Inform Software Corp., Chicago (1996).
[3] von Altrock, C., Gebhardt, J., "Recent Successful Fuzzy Logic Applications in Industrial Automation"
[4] von Altrock, C., "Fuzzy Logic & Neuro Fuzzy Applications Explained", Prentice Hall, ISBN 0-13-368465-2
[5] Zimmermann, H.-J., "Fuzzy Set Theory -- and its applications", Second Revised Edition (1991), Boston, Dordrecht, London, ISBN 0-7923