N.A. Borsuk – Ph.D.(Eng.), Associate Professor,  Department «Information Systems and Networks», Kaluga branch of the Bauman MSTU

E-mail: borsuk.65@yandex.ru

E.O. Deryugina – Ph.D.(Eng.), Associate Professor,  Department «Information Systems and Networks», Kaluga branch of the Bauman MSTU

E-mail: syvorova_eo@mail.ru

S.M. Latsin – Student,  Department «Information Systems and Networks», Kaluga branch of the Bauman MSTU E-mail: semen.vagon@mail.ru

Ya.V. Ryabtsev – Design Engineer,  Department of Innovation Projecting, JSC «Typhoon» (Kaluga) E-mail: YaroslavRyabtsev@yandex.ru

An important feature of the use of power converters on various transports is the limited availability of on-Board energy. This encourages developers to look for ways to improve the efficiency of such converters, taking into account the growth of their power consumption and the requirements to reduce their size. To improve the efficiency of energy converters, while maintaining or reducing the size of its components, developers are forced to resort to the use of more complex physical principles of their construction. Resonant pulse LLC-converters are widely used in modern power supplies. The circuitry of these converters is best manifested itself when working at high power with a small number of elements. The article considers one of the possible variants of the resonant pulse Converter circuit, namely the full-bridge LLC-Converter. A model of the Converter is built in the Matlab environment. Based on the results of experiments with the model, a graph of the gain of the Converter from the switching frequency and showing the behavior of the Converter at different loads is constructed. The problem of control of the resonant transducer is determined by the nonlinearity of its characteristics in the vicinity of the resonant frequency. Because of this, at high loads, when the Converter is forced to operate at frequencies close to the resonance, its transfer characteristic is constantly changing. To solve this problem, it is proposed to build a control system of the Converter as a system with closed circuits of self-adjustment. The article proposes to introduce an artificial neural network and a digital filter into the corrective device of the Converter control system. Neural network and digital filter are implemented at the software level in a specialized microcontroller.

The input of the neural network receives the values of the output parameters of the voltage Converter, its reaction will be the changed values of the coefficients of the transfer function of the digital filter of the correcting device. The new values of the filter coefficients change the position of the zeros and poles of its transfer function, as well as its gain, which leads to the stabilization of the Converter. The neural network has 26 neurons, input, output and two hidden layers. To train the neural network, a large number of correcting devices of the Converter were developed and experiments were carried out under different input conditions of its operation. The experimental data allowed the drawing of the training sample. One example of a training sample is given. The network is trained by specialized tools of the Matlab package and has a small standard error. The operation of the Converter with an adaptive control system is tested on the model built by means of the «Matlab» package. The response of the adaptive system to a single impact is smooth, without over-regulation of the transient response. The duration of the transition process was 60 milliseconds, which in the conditions under consideration is a very satisfactory result. The developed model of the electric power Converter with an adaptive control system based on a neural network shows its superiority over converters with a conventional controller. Further improvement of the work can be developed in the direction of improving neural network models and digital filters of corrective devices.the construction of the control system of the Converter as a system with closed circuits of self-adjustment figure.

1. Vorobev S. Postroenie sistemy elektropitaniya po printsipu IBA dlya vysokonadezhnoi radioapparatury. Pochemu by i net?. Komponenty i tekhnologii. 2015. № 6. S. 23−31.
2. Vorobev S. Moduli elektropitaniya SynQor s rasshirennym funktsionalom. Silovaya elektronika. 2015. № 5. S. 45−51.
3. Deryugina E.O., Novikov R.N., Ryabtsev Ya.V. Metody borby s pomekhami vkhodnykh tsepei preobrazovatelei napryazheniya. Elektromagnitnye volny i elektronnye sistemy. 2017. T. 22. № 3. S. 11−16.
4. Web-portal dlya razrabotchikov elektroniki. Rezonansnye LLC–preobrazovateli. Chast vtoraya: ot pryamougolnykh impulsov k sinusoidalnym signalam. URL = https://www.terraelectronica.ru/news/5322 (data obrashcheniya 01.03.2019).
5. Shraiber G. 300 skhem istochnikov pitaniya. M.: Mir. 2010. 173 s.
6. Kulik V.D. Silovaya elektronika. Avtonomnye invertory, aktivnye preobrazovateli. SPb.: SPbGTURP. 2010. 90 s.
7. Besekerskii V.A., Popov E.P. Teoriya sistem avtomaticheskogo regulirovaniya. Izd. 3-e, ispr. M.: Nauka. 1975. 768 s.
8. Spravochnik po teorii avtomaticheskogo upravleniya. Pod red. A.A. Krassovskogo. M.: Nauka. 1987. 712 s.
9. Komartsova L.G., Maksimov A.V. Neirokompyutery. Izd. 2-e, dopoln. M.: Izd-vo MGTU im. N.E. Baumana. 2004. 320 s.
10. Syuzev V.V. Osnovy teorii tsifrovoi obrabotki signalov: Ucheb. posobie. M.: RTSoft. 2014. 752 s.
11. Drach V.E., Korneev A.A., Chukhraev I.V. Modelirovanie elektricheskikh skhem v sovremennykh SAPR. Elektromagnitnye volny i elektronnye sistemy. 2017. T. 22. № 3. S. 36−41.
12. Aliev M.Yu., Maksimov A.V., Tatyanich N.V. Metody otsenki chisla otchetov tsifrovogo filtra po shirine perekhodnoi zony amplitudnochastotnoi kharakteristiki. Elektromagnitnye volny i elektronnye sistemy. 2016. T. 21. № 7. S. 27−31.
13. Deryugina E.O., Ryabtsev Ya.V. Algoritm dvoichnoi uglovoi modulyatsii. Almanakh mirovoi nauki. 2016. № 5−1 (8). S. 72−73.