ИДЕНТИФИКАЦИЯ ПАРАМЕТРОВ СХЕМЫ ЗАМЕЩЕНИЯ АСИНХРОННЫХ ДВИГАТЕЛЕЙ ПРИ ПОМОЩИ НЕЙРОННЫХ СЕТЕЙ

Abstract

Relevance AC electric drives holds a significant share in industry (80–90 %), as they are widely used in various devices and machines of various industrial capacities, ranging from conveyors, pumps and cranes, to drives for metallurgical production mechanisms. This is due to their ease of operation, reliability and high efficiency. The most common adjustable AC electric drive in industry is the «Frequency Converter — Asynchronous Motor» (FC–AM) electric drive. Knowing the exact parameters of the AM in such electric drives makes it possible to optimize their operation, improve efficiency and extend their service life. Also, this helps prevent overloads and increase the reliability of equipment, which in turn reduces operating costs and the accidents risks. With the development of technology, parameter identification is becoming increasingly important for adapting motors to new operating conditions and production requirements. Accurate parameters identification of also makes it possible to detect and predict the motors condition, which reduces the malfunctions risk and increases the equipment service life. Thus, knowledge of the asynchronous electric motors parameters plays a key role in increasing production efficiency and reducing operating costs. Aim of research The article purpose is to modernize the existing methodology for conducting asynchronous motors tests, regulated by the current State Standard 7217-87, through the introduction of neural networks for further identification of parameters based on them. The article provides a rationale for choosing the neural networks structure, its training algorithms, and also experimentally selected the hidden layer optimal configuration for each experiment. Neural networks were trained on motors with a power of 10-100 kW with a rated voltage of 380 V, a frequency of 50 Hz, and a synchronous rotation speed of 1500 rpm. Research methods The studies were carried out using Matlab Simulink software, data processing was carried out in Matlab and MS Excel software. The experiments, regulated by the current State Standard 7217-87, were carried out on computer models. Among the research methods used in the article, we can distinguish such as experiment (simulation in accrdance with State Standard 7217-87 measuring experiments), measurement (obtaining output parameters of the AM current strength in a computer model), comparison (comparing the results of training neural networks using built-in metrics tools). Results Neural networks have been developed to identify asynchronous motors parameters that process the experiments results in the current State Standard 7217-87. Also, the creating neural networks methodics describe in the article can be used to create similar neural networks for other types of AM.