Synthesis and modeling of the electric drive control system based on the reference model of the neural regulator

The purpose of this work is to implement intelligent algorithms for the synthesis of control systems for electric drives of power supply systems at water transport facilities using artificial neural networks. The use of such intelligent algorithms will make it possible to carry out in practice the digital transformation of hardware units of regulators (controllers) in control systems for various objects, including electric drives, into mathematical algorithms based on neural network controllers. Such controllers, for example, those using a reference model, are more preferable when controlling nonlinear objects, since the neural networks on which they are based are nonlinear. In view of this, the scope of their application has been significantly expanded in the further development of methods for computer monitoring and parametric identification of ship and shore power supply management models, as well as the analysis and forecasting of energy efficiency indicators of their operating modes. The procedure for synthesizing a neural network regulator built on the basis of a reference model to stabilize the angular velocity of a DC motor is considered, aiming to compensate for oscillations occurring in the drive control loop. Using a PID tuner, the parameters of the PID controller were determined, significantly affecting control quality and allowing it, together with a typical first-order astatic link, to perform the function of a reference regulator for training the neural network controller. It is shown that the selected parameters of the neural model of the controlled object and the neural network reference regulator made it possible to significantly improve the quality of the transient process and eliminate oscillations in the DC motor drive control. The indicators and characteristics of the training quality of the neural network regulator and the neural model of the object with the selected training parameters are presented. An algorithm for training the neural model of the controlled object and the neural network regulator based on the dynamic nature of the backpropagation of error deviations of output signal values from reference ones in a multilayer neural network is proposed, with the purpose of correction through introducing adjustments to the values of synaptic weight coefficients. The algorithm can be applied in control systems for electric drives of unmanned objects, including aircraft, waterborne, and land-based systems in inland water transport.

1. Baryshnikov, S. O., D. V. Dmitrienko, V. V. Sakharov and A. A. Chertkov. Modeli i algoritmy upravleniya ob”ektami vodnogo transporta v usloviyakh tsifrovoy transformatsii SPb: Izd. «Zanevskaya ploschad’», 2022: 520.

2. Sakharov, V. V., A. A. Kuz’min and A. A. Chertkov. Modeli i algoritmy optimizatsii tekhnologicheskikh protsessov na ob”ektakh vodnogo transporta v srede MatLab Sankt-Peterburg: Federal’noe gosudarstvennoe byudzhetnoe obrazovatel’noe uchrezhdenie vysshego obrazovaniya Gosudarstvennyy universitet morskogo i rechnogo flota im. admirala S. O. Makarova, 2015: 436.

3. Chertkov, A. A., C. V. Saburov and Ya. N. Kask. “Algorithm for identifying parameters of production function models using a neural network.” Vestnik gosudarstvennogo universiteta morskogo i rechnogo flota im. Admirala S. O. Makarova 15.6 (2023): 1096–1104. DOI: 10.21821/2309-5180-2023-15-6-1096-1104.

4. Tsvenger, I. G. and I. R. Nizamov. “Primenenie neyrosetevykh regulyatorov v sistemakh upravleniya elektroprivodami.” Vestnik Tekhnologicheskogo universiteta 20.8 (2017): 111–114.

5. Chertkov, A. A., D. S. Tormashev and S. V. Saburov. “Dynamic systems parametric arrange by means pid tuner in matlab.” Vestnik gosudarstvennogo universiteta morskogo i rechnogo flota im. admirala S. O. Makarova 5(27) (2014): 164–171.

6. Khaikin, Simon. Neural Networks: A Complete Course. 2nd Edition. with anrl. Moscow: Williams Publishing House, 2006: 1104.

7. Vakulenko, S. A. and A. A. Zhikhareva. Prakticheskiy kurs po neyronnym setyam. SPb.: Universitet ITMO, 2018: 71.

8. Bobtsov A. A., V. V. Grigoriev, and P. V. Karev et al. Actuating devices and systems for micromovements. St. Petersburg: ITMO University, 2017: 134.

9. Davidov B. Automatic optimization of Simulink models and controllers Portal of scientific and practical publications Web. 1 Mar. 2025 <https://portalnp.snauka.ru/2013/10/1194>

10. Belov, M. P., I. S. Nosirov and A. M. Belov. “Synthesis neural network controller for two-mass electromechanical system of lathe machine’s feed drive.” Izvestiya SPbGETU LETI 8 (2018): 70–76.

11. Emel’yanov, A. A., V. V. Beskletkin and I. M. Agzamov et al. “Matematicheskoe modelirovanie dvigatelya postoyannogo toka v sisteme otnositel’nykh edinits v Matlab i Si.” Molodoy uchenyy 11(249) (2019): 1–7.

12. Wang, L. PID Control System Design and Automatic Tuning using MATLAB/SimulinkJohn Wiley & Sons, Ltd, 2020: 344. DOI: 10.1002/9781119469414.index.