FORECASTING THE AVERAGE MONTHLY WIND SPEED WITH THE USE OF THE BAYES APPROACH TO FORECASTING

Abstract

The article is devoted to the interpretation of the classical Bayesian approach to solving prob- lems in the field of alternative energy, namely forecasting the average monthly wind speed, taking into ac- count expert opinion. The proposed interpreted Bayesian approach to forecasting allows, with a small frac- tion of the error, to determine the forecast value of the average monthly wind speed, taking into account the expert's amendment, that is, to fulfill the forecast taking into account the knowledge of the meteorologist. At the same time, the interpreted method repeats all stages of the orthodox scheme of the forecasting process.
As a result of the research and interpretation, the method of calculating the forecast value of the average monthly wind speed in January 2018 based on the January data for the period from 2012 to 2017 is realized. Moreover, the method allows not only to take into account the value of the wind speed in past periods of measurements, but also the seasonal change in the nature of the wind and its stochastic nature. The main advantage of the interpreted method is that the predicted wind speed forecast for January 2018, using the Bayesian approach to forecasting, repeats the actual average monthly value with an accuracy of 96.5%. This suggests that this approach is much more accurate than systems based on neural networks (prediction accu- racy of 70-80%) or using the Boxx-Jenkins methodology (integrated model of autoregression) with a predic- tion accuracy of only 60-70% ( using forecast only using a time series of wind speed, taking into account seasonality).