MODEL OF SHORT-TERM FORECAST OF ELECTRICAL ENERGY CONSUMPTION OF URAL UNITED POWER SYSTEM BY SEPARATING OF A MAXIMAL SIMILARITY SAMPLE INTO THE POSITIVE AND NEGATIVE LEVELS

Abstract

The article considers a model to forecast electrical energy consumption on the basis of the forecasting algorithm with the division of the maximal similarity sample into positive and negative levels with different approximation equations for positive and negative values. Model is tested with actual daily data of United Energy System of the Wholesale Electrical Energy and Power Market in Russia. Namely, we use Ural United Energy System data from 2009 to 2015. Based on the proposed algorithm, a forecast for the first five days of January 2016 is obtained. A forecasting error achieves 0.98 % during test. We substantiate and describe step by step an algorithm to construct a model to forecast a time series. Such algorithm is more stable for stationary series. Therefore pre-time we bring a
time series of electrical energy consumption to a stationary form. To this end, we find variances of the original time series of electrical energy consumption. The proposed scientific tools are recommended in the operating activity of electrical energy subjects for short-term forecast of the basic parameters of the
Energy Market to reduce the penalties by improving the forecasting accuracy.