In this paper, Artificial Neural Network (ANN) technique has been used for the estimation of voltage THD (Total Harmonic Distortion), and current THD, mainly from input and output measurements of five different chorded induction motors fed from a pulse width modulation inverter voltage supply. A sinusoidal pulsewidth modulation (SPWM) inverter feeding five different chorded three-phase induction motors were tested up to first thirty harmonic voltage component at different loads. The results show that the artificial neural network model produces reliable estimates of voltage THD and current THD .
I.INTRODUCTION ANN models have been applied to a large number of problems because of their non-linear system modeling capability by learning ability using collected data. They offer highly parallel, adaptive models that can be trained by the experience. In fact, ANN models have the universal approximation property that means under mild conditions on the data, they can fit any data set with an arbitrary high precision, provided that there are a sufficient number of parameters in the model. However, when there are too many parameters compared to the number of data available, the over fitting phenomenon appears. The known data used for training are well fitted, but the function has no sense between points used for training [1].During the last decade ANN models have been applied widely to prediction of the data. Such a prediction study has been completed in this paper, to compare the effectiveness of artificial intelligence approach. A two layer feed forward neural network trained by the back propagation technique employed in the stator voltage THD estimation. Therefore , a sinusoidal pulse-width modulation (SPWM) inverter feeding five different chorded three-phase induction motors were tested up to first thirty harmonic voltage component at different loads and different switching frequencies up to 15Khz. The number of all measurements results obtained from experiments are 4369. 15% of this data were used for validation, 15% were used for test and 70% were used for training the neural network. Based on experimental results, the artificial neural
