Abstract The article discusses the improvement of thermal working conditions in underground mine workings, using local refrigeration systems. It considers the efficiency of air cooling with direct action air compression refrigerator of the TS-300B type. As a result of a failure to meet the required operating conditions of the aforementioned air cooling system, frequently there are discrepancies between the predicted (and thus the expected) effects of its work and the reality. Therefore, to improve the operating efficiency of this system, in terms of effective use of the evaporator cooling capacity, quality criteria were developed, which are easy in practical application. They were obtained in the form of statistical models, describing the effect of independent variables, i.e. the parameters of the inlet air to the evaporator (temperature, humidity and volumetric flow rate), as well as the parameters of the water cooling the condenser (temperature and volumetric flow rate), on the thermal power of air cooler, treated as the dependent variable. Statistical equations describing the performance of the analyzed air cooling system were determined, based on the linear and nonlinear multiple regression. The obtained functions were modified by changing the values of the coefficients in the case of linear regression, and of the coefficients and exponents in the case of non-linear regression, with the independent variables. As a result, functions were obtained, which were more convenient in practical applications. Using classical statistics methods, the quality of fitting the regression function to the experimental data was evaluated. Also, the values of the evaporator thermal power of the refrigerator, which were obtained on the basis of the measured air parameters, were compared with the calculated ones, by using the obtained regression functions. These statistical models were built on the basis of the results of measurements in different operating conditions of the TS-300B refrigerator, both on the test stand in the manufacturer’s laboratory and in the workings of underground mines. The evaluation of the measurement data distributions, as well as an analysis of the basic descriptive statistics of the mentioned variables were carried out, determining their measures of central tendency, location, dispersion and asymmetry.