Open AccessNeural network architecture choice for modelling various configurations power system
Author(s) -
Б. В. Кавалеров,
G. A. Kilin,
A. I. Suslov
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1886/1/012007
Subject(s) - intellectualization , artificial neural network , computer science , power (physics) , control engineering , reliability (semiconductor) , network architecture , turbine , electric power system , reliability engineering , engineering , artificial intelligence , mechanical engineering , psychology , physics , computer security , quantum mechanics , psychotherapist
Gas turbine power plants based on converted aircraft engines have been successfully produced in Perm city since 1991. Their power range is steadily expanding. Such gas turbine electro power stations have significant advantages. However, the need to make full use of these advantages requires power station control processes significant intellectualization. Intellectualization will provide an opportunity to improve the electricity quality, increase the power supply reliability and improve environmental friendliness. To achieve these goals, it is necessary to study the behavior of gas turbine power electro station in different operation modes and in various configurations power system. Such research can only be carried out on the basis of mathematical models. Moreover, these models should be high-speed, since the research of control algorithms requires a large number of various experiments in a limited time. One of the promising ways to obtain such mathematical models is to use the artificial neural networks apparatus. To construct a gas turbine power electro station mathematical model, it is necessary to choose such a neural network architecture that will allow obtaining a model for an acceptable period of time and will allow modeling the processes with the required accuracy. Therefore, the problem of choosing the neural network architecture is an urgent research priority. On the basis of the performed studies, the architecture of a neural network for modeling various configurations power system is proposed and substantiated.