Neural-Network-Based Process Controller Design and On-Line Application to Fluidised Bed Combustion

摘要

Intelligent on-line control has been increasingly popular in the past decade. In this paper, a neural-network-based controller utilising a modified error term in the backpropagation algorithm is presented for the purpose of on-line control of the nonlinear time-varying fluidised bed combustion (FBC) process. The aim of the modification of the error term was to improve the controller performance by enabling the neural network to perform a 'negative hysteresis' action. The general design steps, alterations in the controller and performance tests were initially carried out on a simulation model of the process. It was observed that the proposed controller is successful in accomplishing the bed temperature control of the FBC process in the absence of any human operator. Furthermore, performance tests made on the simulation model showed that without the presence of offline pre-training, the proposed controller performs better than the conventional neurocontroller in convergence time and overshoot.