ON-LINE TRAINING MODELLING AND CONTROL USING NEURAL NETWORKS, FOR VAPOUR PRESSURE REFRIGERATION SYSTEMS

摘要

This work uses the Neural Network (NN) technique to model and control an experimental vapor compression refrigeration system set up, using two manipulated variables (compressor rotation, expansion valve electronically actuated). It was used a dynamic perception multi layed NN, with the NN delayed, and a back propagation algorithm with sigmoid actuation fraction. To model the system and a Neural prediction with correction, MISO type control. Paper relates, training techniques used using single and simultaneous controlled variables inputs the NN technique used shows good dynamic stability, short time actuation and easy way of implementation.