Improved Process Control in Coal Preparation

The objective of this project was to develop and evaluate new control algorithms more appropriate to the needs of coal preparation plants than existing PI(D) algorithms.

Whiten's Form Generating Controller (FGC) which automates the techniques used in manual control was chosen for investigation. The FGC generates a time sequence of changes to the process input while watching (through the use of a simple process model) to ensure that the process output is responding correctly.

The major advantages of the FGC are:

• the form of the control can be pre-defined, making tuning much easier,
  
• its ability to deal with the long dead times often found in process plants as a result of transport delays.

Saraji Coal Preparation plant was chosen as the site to evaluate the FGC. It has a standard DSM medium density and sump level control system using PID loops. These loops interact strongly, making good control difficult.

A dynamic model of the Saraji sump level and medium density control system was developed and validated. The FGC controller was applied to the model with excellent results, the performance of both sump level and density control was better than with PI controllers. The results looked extremely promising for the plant trails.

However Saraji were unable to continue with the project and the work was transferred to Goonyella, which has a similar control system. Site work was carried out at Goonyella in June and October 2000.

Results from the FGC in plant application were not as good as expected. While the FGC was capable of controlling the plant, performance was poorer than with the existing PI loops.

Lack of time precluded a detailed investigation of the reasons for the poorer performance. It is possible that the process model need further tuning, and the implementation of the FGC in the UNAC software used at Goonyella may have been incorrect.

It is recommended that a short project be undertaken to complete the evaluation of the FGC, given its excellent performance on the simulated plant.

The dynamic simulation also offers considerable potential for on-line predictions of plant performance, using real plant data as input to the model to keep it tracking correctly.