Neural network-based Load Balancing and Reactive Power Control by Static VAR Compensator

Abstract

unbalanced loads in three phase systems produce undesired negative and zero sequence currents. Negative sequence currents will cause excessive heating of electrical machines, saturation of transformers, ripple in rectifiers or even instability problems of generators. Zero sequence currents cause not only excessive power losses in neutral lines but also protection and interference problems. To achieve a balanced operation for unbalanced AC system and improve power transfer capability, reactive compensation is used. This study use Artificial Neural Network (ANN)deals with an active scheme aiming to attenuate the negative sequence component of the line currents associated with an unbalanced three phase load. The system also assures a unity power factor at the load bus. The proposed structure uses a Y-connected capacitor bank and _-connected thyristor reactor which is controlled by ANN, So that, the amount of reactive power is being controlled. Three different networks are learned by both real and imaginary part for three phase load currents (Ix, Iy, Iz). Hundreds of unbalanced cases in a 33 (KV) system are simulated for this purpose. Study show that ANN-based reactive power compensation is completely suitable for an on-line load balancing and reactive power compensation in the system by measuring both real and imaginary part of three phase load currents and processing them by ANN.