Mathematical modeling and forecasting of power consumption values based on the decision trees algorithm

The article presents the results of the analysis of average daily values of daily graphs of electricity consumption, air temperature, natural light, cloudiness and precipitation in Moscow. The dependences of electricity consumption on air temperature, cloudiness, precipitation and illumination are obtained. A model is proposed based on machine learning using the algorithm of decision trees of regression taking into account data on air temperature, cloudiness and natural light. This model predicts the values of the time series of power consumption taking into account meteorological factors based on simple decision-making rules derived from the characteristics of the original data. The article presents the results of forecasts of average daily values of power consumption using a forecast model based on the decision tree method for selected periods in April, June and October. The results of forecasting the average daily values of power consumption show that in most test examples the obtained error values are no more than 5 %. In addition the devel-oped algorithm allows using seasonal patterns between daily graphs of power consumption and meteorological factors. Python libraries were used to implement the modeling of mathematical models for analyzing and forecasting time se-ries. Data from the meteorological software and hardware complex PAK "Meteo" were used in the analysis and trans-formation of time series of meteorological factors.