Detection and classification of PQ disturbances in the supply to induction motor using wavelet transform and feedforward neural network

In real world applications, induction motors are normally operated on load. Any variation in the supply voltage, would adversely affect its normal operation. Hence it is important to detect any power quality disturbance and identify the nature of the disturbance. A major issue is to find the motor parameter which can satisfactorily identify and classify the disturbance. This paper proposes a novel method for the detection and classification of PQ disturbances in the supply to induction motor using the signature of the stator current. The tools used for the analysis are wavelet transforms and neural network. In the first stage the stator current is sampled and DWT is applied to it to capture the feature of the disturbance. The DWT co-efficients so obtained are used as input to a feed forward neural network, in the second stage, to classify the disturbance. This method has been tested for different PQ disturbances such as balanced voltage sag, balanced voltage swell, unbalanced voltage sag and unbalanced voltage swell (or simply called as unbalance in the magnitude of the supply). The method has been found to be effective for varying degrees of disturbance. It is observed that the proposed network has performance efficiency greater than 97%.