| Abstract |
It is known that the three-phase squirrel cage induction motors are the major world energy consumers.
Thus, the research is focused on the reduction of the energy consumption field. Because the transient
regimes (desired and accidental) are dominant, in this paper a method of optimal control based on
energetic criteria for dynamic regimes of AC drive with three-phase induction motor, is involved. The
disadvantages of the matrix Riccati differential equation solution (MRDE) are avoided by realizing a
nonrecursive one. The optimal control synthesis consisting of the determination of the stator threephase
currents system, based on the longitudinal and transversal components of the stator phasor
current. The optimal control law provides dynamic regimes with minimal input energy consumption
and minimal windings energy dissipation. The optimal nonrecursive solution is obtained by numerical
integration of the MRDE. Using a feed forward neural network approximates the optimal control
solution. In order to improve the back propagation training algorithm a Recursive Gauss-Newton
Training Algorithm (RGNTA) was presented. Thus, the second derivatives information was involved
in order to provide the optimal solution. The convergence of the RGNTA is faster then the back
propagation one, and more robustness. In order to establish the neural network (NN) architecture the
author had in view to avoid reaching the local minimum [1]. The experimental results show both the
optimal characteristics of the MRDE solution and the properties of the NN. Also, are emphasized the
advantage of this control strategy versus classical control system in AC drives. The rotor field oriented
IM controlled at constant flux was considerate. |
