ESTIMATIONS OF ERROR OF PROGNOSIS MODELS AND PROGNOSES OF THE USED ELECTRIC ENERGY ARE ON OBJECTS OF POWER MARKET

Abstract

The article examines the estimation of models of forecasting of electric energy on industrial
objects of the energy market of Ukraine. The exponential first-order smoothing model, Holt model, Winter model,
Box-Jenkins model and Singular spectrum analysis (SSA) were used in the study. The methods used are simple to
predict electricity and allow you to investigate the behavior of the prediction error, depending on how the model is
estimated. Using modern methods of gathering information will allow you to make the right decisions in real time.
The article can significantly affect the correct choice of both model parameters and method of forecasting electricity.
The research will allow to make the forecasting of electric energy with less error in the objects of the energy market
of Ukraine knowing the statistics of errors of forecasting methods.
The study has shown that estimation of forecasting models will allow to make a choice of a forecasting
model in a short time and the forecasting error will reach up to 10%.
Methods for estimating the error of power consumption forecasting models have shown that the standard
deviation ratio and the Taylor coefficient are the most suitable for comparison purposes because it is a number
between 0 and 1.
The root value of the root mean square error of the RMSPE for the methods is very small, namely for the
methods, on the basis of which it can be stated that the accuracy of the forecasts is high. The mean absolute MAPE
error is for the methods, respectively, which also indicates good accuracy.
The values of the dimensionless obtained Tail coefficients, which are equal and corresponding,
respectively, indicate that the forecasts have high accuracy (forecasts are absolutely accurate if U = 0).
The calculations confirmed that the SSA method is appropriate for use in short-term forecasting tasks and
when planning power consumption modes.
Electricity forecasting is highly dependent on daily and seasonal power consumption. In energy associations
where there is a high irregularity in electricity consumption schedules, and there is a significant deviation of
consumption from the seasonal trend, the prediction error is not significant.
When organizing market power sales and power interconnections with high irregularity in electricity
consumption schedules, the forecasting error will be high and they will be subject to greater penalties for exceeding
or reducing consumption.
The best results from power consumption forecasting are obtained when using specially designed
weathering models.