Dynamic Voltage Restorer Based on Neural Network and Particle Swarm Optimization for Voltage Mitigation

Dynamic Voltage Restorer (DVR) is one of the most widely implemented power devices used to mitigate voltage unbalance in the grid. The performance of DVR largely depends upon its control strategy, in which controller plays an important part. Literature has shown that the commonly used proportional integral (PI) and neural network (NN) controller have many inherent disadvantages including high total harmonic distortion (THD) and high delay time. In this paper, we have replaced the PI controller with a neural controller, whose weights are trained using Particle Swarm Optimization (PSO). A comparative analysis of the DVR performance is done in MATLAB SIMULINK environment for three controllers-PI, NN with back propagation and NN with PSO for 30 and 80 % voltage sag, 40 % voltage swell and unbalanced voltage (1-phi). The results obtained document that the hybrid neural controller with PSO has least distortions and is most robust of the three.