Vector control of an induction motor has advantage of a quick torque response, and has been applied to various industrial applications.
In the design of a speed control system of induction motors, PI controller has been widely used because its structure is very simple. However, it is difficult to obtain robust and stable speed control characteristics because the gain of the controller can not be adjusted automatically when the load disturbance or system parameter change.
The motor used in a reciprocating air compressor experience abrupt load change by the movement of piston. So, its speed is fluctuated.
This study proposes a new adaptive control system with conventional vector controller for a reciprocating air compressor. The proposed system consists of a load torque observer and a feed-forward compensation using neural network to obtain a robust speed control characteristic. The observer is designed based on the Gopinath theory. And the neural estimator is consisted of two layers, and is used to provide a real-time adaptive estimation of motor dynamics. The LMS(Least Mean Square) algorithm which has widely been used is applied as the learning algorithm for this network to minimize the difference between the actual speed and the predicted value by the neural estimator.
To verify the effectiveness of this algorithm, a computer simulation and a experimental test are carried out on the basis of theoretical consideration.
From the experimental result, it is confirmed that the transient responses under the 1[atm] and 2[atm] condition are reduced 80[ms] and 50[ms] respectively compared with conventional method, and also steady state speed fluctuations are reduced 50 [rpm] and 80[rpm].