ANALYSIS OF METHODS AND MODELS FOR AUTOMATION OF ENERGY CARRIER CONSUMPTION FORECASTING

Abstract

The usage of statistical models, artificial neural networks, and adaptive neuro-fuzzy systems for forecasting energy carrier consumption is considered in the article. It is shown that the use of artificial neural networks makes it possible to model complex dependencies between input and output parameters. Disadvantages of the use of neural networks for forecasting gas consumption are highlighted, in particular, a large amount of data for training the network, availability of expert knowledge and experience of the researcher, high computational complexity of training and application of the network. The advantages of mathematical models for forecasting the volumes of gas consumption are also determined, which allows to obtain more accurate results, provides the possibility of taking into account expert experience, allows to adapt the model to changing conditions and establish cause-and-effect relationships. The feasibility of developing the automated system for forecasting natural gas consumption volumes based on a combined forecasting method is substantiated. The combined method should combine the deterministic dependence of consumption volumes on the main input parameters and the use of a neural network to describe the influence of many random factors.