Neural network based self-tuning control of a switched reluctance motor drive to maximize torque per ampere

Abstract

On-line self-tuning control is essential to optimize the performance of a Switched Reluctance Motor (SRM) Drive in the presence of parameter variations. This paper introduces an advanced adaptive Neural Network (NN) based control to maximize torque per ampere in the low speed re@on. The proposed control technique utilizes a heuristic search method to find the change in the optimal excitation instances in case of parameter variations. Based on the results of this heuristic search, the NN employs an incremental learning to adapt its network weights. Computer simulations are performed to verify the applicability of the proposed algorithm. Experimental results are provided to demonstrate the working of the self-tuning controller.

A New Fuzzy Logic based Modeling and Simulation of a Switched Reluctance Motor

“…Therefore, more advanced control techniques need to be used which will minimize the noise and ripples in torque. Hence, a fuzzy logic and neural network [31] based controller is studied and a fuzzy based controller is proposed. The operation of a fuzzy controller is split into three stages: Justification of crisp input values, fuzzy inference using a knowledge base, and defuzzification of the result of the inference process to give crisp output values, which is used for controlling.…”

“…The application of sensors for rotor position is not viable all the time, therefore various schemes for sensorless position sensing based on static and real time data is developed [10]- [11]. Methods including ANN [12] and Fuzzy logic [13] are now widely used. The benefits of indirect or sensorless sensing are found to be the elimination of electrical connections of the sensor, reduced size, less maintenance, insusceptible to environmental factors and increased reliability.…”

Intelligent tuning of commutation for maximum torque capability of a switched reluctance motor

“…As to the self-tuning technique presented in [13], a simple algorithm along with a lookup table is used to optimize the TPA with additional computer simulation proving its existence and uniqueness. Concerning the optimization of the TPA made in [14] and [15], the artificial neural networks (ANNs) are applied to perform the online self-tuning of turn-on or turn-off angle. However, a lot of simulated or measured data are required for training and establishing the NNs.…”