Time Scale Control Using Dynamic GMDH Neural Network Forecasting Based on Real Measurement Data

This article presents the results of the conducted research work related to the dynamic forecasting of the difference values for the Polish Time Scale UTC(PL) for real measurement data, prepared in the form of the time series TS1 and TS2. For the presented time period (the whole year of 2024), the differences between the UTC(PL) and UTC does not exceed ±4.4 ns. The analogous differences for the interval exceeding 2 years are within the range of ±5 ns. Additionally, the obtained forecast results for the last day of forecasting in a given week are very consistent with the forecast results for the first day of the new forecasting week, which illustrates the very good quality of the forecasting and the universality of the forecasting procedure developed by the author using the GMDH-type neural network.