Pre-forecast modeling of airport electricity consumption time series

Abstract

The article analyzes the relevance of pre-forecast modeling of time series of electricity consumption by airports, systematizes the methods and ways of the specified pre-forecast modeling and considers some problems arising in the process of their use. A separate stage of preforecast modeling of electricity consumption by the airport is proposed, which contributes, on the one hand, to a fairly quick receipt of primary information about the forecasted object, and on the other hand - to a more effective and adequate final forecast. It is proposed to build a series of neural network models at the stage of preliminary forecasting, including convolutional, recurrence. As a model example, a neural network preforecast model of electricity consumption for the Lviv International Airport is built on the basis of statistical data for the period of relatively stable development of the Ukrainian economy. A comparative analysis of the obtained results of the neural network model with the constructed trend-seasonal model using analytical methods was carried out, which gave a positive result. Conclusions are made on the prospects of building preforecast models of time series of electricity consumption by the airport using neural networks.