Application of a Model-Based Predictive Controller on a Vacuum Heat-Exchange Network for the Bayer Process

"This paper describes the implementation of model-based predictive control (MPC) on a vacuum heat exchange network at Alcan's Jonquière Complex. The plant produces alumina from bauxite ore using the Bayer Process.The MPC system was commissioned in 2000, by Alcan personnel and an Advanced Control Engineer from Matrikon. The application has achieved an uptime of about 97%.The benefits were:• Energy savings : 0.5% - live steam reduction• Reduced variation: standard deviation of key temperatures reduced by 50%• Plant throughput increase: via the automated use of surge tanks• Decreased supervision time: by plant operations• Faster and easier diagnosis of problems and plant constraintsThe benefits were higher than expected with a payback time around 12 months.The purpose of the vacuum heat exchange network is to recover energy from a hot stream (pregnant liquor) by transferring it to a cold stream (spent liquor). The system is very tightly coupled, and the main constraints of the multivariable control system are: • to control the many tank levels involved in the process (8 tanks)• to maximize the temperature of the heated streams (spent liquor)• to prevent cooled streams (pregnant liquor) from cooling to below a minimum temperatureThe system includes 9 manipulated, 14 controlled and 2 feedforward variables.The advanced control implementation runs under an expert system environment and provides added flexibility to the overall control strategy."