Reliable Multi-Objective On-line Re-Optimisation Control of Batch Processes Based on Bootstrap Aggregated Neural Networks

Abstract

The paper presents a reliable multi-objective re-optimisation control strategy for batch processes based on bootstrap aggregated neural networks. Bootstrap aggregated neural networks not only give better generalisation performance than single neural networks but also provide model prediction confidence bounds. In order to overcome the problem of unknown disturbances, on-line re-optimisation is carried out to amend the control policy for the remaining batch duration. In addition to the process operation objectives, the reliability of model prediction is incorporated in multi-objective optimisation in order to improve the reliability of the obtained optimal control policy. The standard error of the individual neural network predictions is taken as the indication of model prediction reliability. The proposed method is demonstrated on a simulated fed-batch process.