Automatização e controle inteligente on-line de sistemas de refrigeração utilizando redes neurais artificiais

Abstract

Usage of smart Control Systems in industrial processes has quickly increased on last decodes, principally in hard mathematical modelling systems. Systems efficiency of refrigeration cycles is directly related to the capacity of maintaining temperatures and pressure to the need of process. Condensing and evaporating temperature are very important for the energy consumption and the system performance, specially when it has been influenced by external factors. This thesis searches for some vantages about refrigeration systems using variable speed compressor and an electronically driven expansion valve. Influence of variable speed and opening/closing valve, in order to reduce the energy consumption, are handled using an identification method and systems control. This method is Black-Box model based on Artificial Neural Networks. Tests were initially developed to analyse the dynamical behaviour in a refrigeration system prototype which was assembled in the Energy and Thermal Systems Laboratory (LEST) at UFU; in order to create a data base to feed the neural network. A methodology of experimental planning was used, to analyse the individual sensibility with the variable interaction and the systems parameters. The control model has as purpose the rational use of energy of this equipment and presents resulted satisfactory.