Method of controlling the condenser irrigation system of a refrigeration plant located in a hockey stadium using a neural network controller

Background. The need to improve the energy efficiency and reliability of refrigeration systems at sports facilities stipulates the need to improve the methods for controlling the condenser irrigation process. The main limitations of existing approaches are related to the lack of flexibility and the complexity of fine parameter calibration. The main objective of the research is to create and implement a neural network controller capable of promptly responding to the current values of environmental variables such as atmospheric temperature, relative humidity and the level of thermal load on the refrigeration plant. Materials and methods. A comprehensive comparative analysis of known cooling control algorithms was carried out, revealing the characteristic problems of the traditional approach. The implemented neural network architecture is based on the principles of Generative Adversarial Networks (GAN), which was trained on an empirical set of historical data collected from operating industrial equipment. Practical verification of the effectiveness of the developed model was carried out by means of a field experiment at a functioning refrigeration plant of a large sports facility, accompanied by detailed monitoring of the climatic situation and thermophysical properties of the refrigerant. Results and conclusions. A statistically reliable increase in the stability of the maintained technological parameters and a noticeable reduction in the power consumption of the cooling system have been experimentally confirmed. The delay time of controller reactivity to random perturbations is reduced, which contributes to the improvement of dynamics and adaptability of the cooling mode to changing operating modes. The application of neural network technologies opens up prospects for optimisation of energy saving and increase of cooling units performance at specialised objects of mass use. The effectiveness of the proposed solution confirms the expediency of its further implementation on similar technical complexes.