ELIMINATION OF POWER QUALITY PROBLEMS IN DISTRIBUTION SYSTEMS

Abstract

In this work the active power line conditioner {APLC) is simulated by using a two stage neural network controller for compensating load current harmonics, load imbalance and power factor. In the first stage the adaptive neural networks is used to obtain the fundamental voltage and current components by continuous training for each samples of voltage and current waveforms. The training is implemented by using the modified Widrow-Hoff (W-H) algorithm. After getting the current samples a simple algorithm is used to find the reference currents. Second stage is the hysteresis neural controller which is used to generate the firing pulses to the IGBTs. Backpropagation technique with Levenberg-Marquardt algorithm is used for training. Based on the same algorithm a new control is implemented for series active filter for compensating the load voltage interruptions. In this the adaptive neural networks is used to obtain the fundamental voltage and current components by continuous training for each samples of current waveforms and trained for voltage waveforms if there a change in system voltage. All the control algorithms are written in MATLAB and the compensated systems are designed in PSCAD. The MATLAB is interfaced to the PSCAD by using a FORTRAN subroutine.