DC electric drive adaptive control system development using neural tuner

Abstract

An adaptive control system development for DC electric motor implemented as “thyristor converter — DC motor” system is considered. Linear controllers based on P- and PI-algorithms are usually used for such plants. So adaptation is decided to be made by these controllers parameters on-line adjustment with the help of a neural tuner. This method allows to keep required transients quality for both armature current and speed control loops under the conditions of both drive electromagnetic and electromechanical time constants values drift. Its usage also does not demand to change existing control system significantly. Mentioned controllers parameters are tuned in accordance with a rule base containing process automation engineer experience. Such rule base defines learning rate values during each neural tuner call. They are used for the neural network on-line training with the help of the backpropagation method. The neural network outputs are K P and K I values. The neural tuner is applied to the mentioned DC motor physical analogue. Linear controllers parameters on-line tuning for the DC drive under consideration allows to improve plant energy efficiency by 2%.