Electric drive control systems with neural network technologies

The research is devoted to the development of hybrid sensorless control systems based on non-adaptive controllers and neural network controllers of various structures. The main methods used in the study are related to the mathematical modeling of these systems in Matlab Simulink. The results of the study are hybrid systems that take into account the heating of the motor windings during its operation, which showed an acceptable processing of the speed signal. The paper proposes algorithms of neural network technologies as a variant of implementing the speed observer structure. The modeling of observers carried out in the study with their subsequent testing in simulation modes made it possible to evaluate the qualities of each of the systems and draw a conclusion about the qualitative superiority of neural network algorithms over the classical mathematical apparatus. The structure of the control system of an asynchronous electric drive with a digital speed monitor is proposed, which, in addition to determining the speed, takes into account the state data of the drive and monitors the state of the engine components. The possibilities of applied implementation of the obtained control systems at various levels of automation are considered: from a simple controller at the drive level to remote cloud spaces. The study of the functioning of the observers in question was carried out by analyzing the dynamics of the speed change at resistance values of 1.25 of the nominal value. The error error in speed when using neural network technologies lies in the range of 0.1•10-6⋯ 0.2•10-6 rad/s.