Vol. 126, No. 4 * Pages 425–582 * October - December 2022

Meteorologists keep searching and running models to provide the most accurate forecast of wind speed in addition to gaining a more detailed understanding of the wind conditions in Hungary. Wind speed and wind energy estimates, forecasts, and their verification are based on wind statistics from a longer or shorter previous period. Consequently, in addition to dynamic methods, purely statistical models also play an important role, i.e., findings that can be obtained from the statistical analysis of the existing database of measured data. The successive phases of the statistical method for producing scientific or operational information that can be extracted from measured, corrected, and stored meteorological data are generally: statistical analysis/processing, creating, verification, and application of the model, recording of the required information. The targeted information in this paper is the daily average of hourly wind speeds. The exact average of this time series can only be determined after the last measurement. To estimate this average during the day, however, the so-called sliding average model has been developed, which can be applied to any climatic element if its measured values are recorded at regular times over a certain period of time. The results presented in this paper are recommended for the preparation of the so-called "timetable", which is one of the most difficult problems for wind farm operators. This is basically the estimation of the amount of electricity produced the following day over short periods. It would be a significant help in the above if we can determine the probability of a decrease or increase in the average wind speed on the next day (and with it, the average daily wind power), or which of these two probabilities is greater. This requires an estimate of average wind speed of the next day. In addition, the results of one of our previous studies on the statistical structure of day-to-day changes in average daily wind speeds were also used. According to the results of the monthly testing of the model over a given period, the frequency of good estimates is between 80.6% and 54.8%.