PREDICTION OF ENERGY EFFICIENCY OF SCREW COMPRESSORS IN AMMONIA REFRIGERATION SYSTEMS WITH FREQUENCY CONTROL USING A NEURAL NETWORK

This paper presents the results of a study on the energy efficiency of screw compressors in ammonia refrigeration systems equipped with frequency control. The focus is placed on the development of a neural network-based model for predicting energy consumption using experimental data collected from operational industrial facilities. The proposed approach accounts for a wide range of operating conditions, including variable load profiles, ambient temperature fluctuations, heat exchange performance, and specifics of automated control systems. The application of an artificial neural network (ANN) enabled high prediction accuracy: for compressors with frequency control, the mean squared error (MSE) was 63.065 and the coefficient of determination (R²) reached 0.992; for compressors without frequency control, the respective values were MSE: 266.231 and R²: 0.978. These results demonstrate the advantage of frequency control in both modeling precision and energy performance. The study confirms that frequency regulation not only reduces energy consumption but also enhances system reliability, equipment longevity, and lowers operational costs. The proposed methodology can be adapted for other types of industrial refrigeration units, opening broad opportunities for further research and the implementation of energy-saving technologies across various industrial sectors. Keywords: energy efficiency, screw compressors, frequency converters, ammonia refrigeration units, optimization, electricity consumption, аrtificial neural network (ANN).