The paper contains the results of investigation of effectiveness of application of various models oriented towards interval time series (ITS) to forecasting behavior of gas distribution networks (GDS) parameters involving the real data obtained during the process of their continuous control. The necessity of taking into account the factor of seasonality caused by periodic fluctuations in the level of the corresponding variable is justified. A comparative analysis of the properties of the special interval modification of the model based on exponential smoothing, neural network and hybrid prediction models in relation to the ITS with seasonality is performed, their merits and demerits are noted.
Keywords: interval-valued time series, exponential smoothing model, neural model, long short-term memory, hybrid model