Robustness of unified power quality conditioner by neural network based on admittance estimation

Abstract This paper attempts to improve the dynamic performance of unified power quality conditioner (UPQC) by using of neural network based on admittance estimation strategy (NN-ADES). For improve the efficiency, tracking capability and robustness of UPQC, a control strategy is used for estimation of fundamental admittance components (conductance and susceptance) of distorted source voltage. In this method, the admittance components weight values of estimated admittance components input conductance clustered value which is very near to weight values of actual admittance components. The NN-ADES strategy is suitable for where load periodicity not constant. The neural network admittance estimation strategy for UPQC is implemented using MATLAB and FPGA board for mitigating balanced or unbalanced load conditions, voltage sag or voltage swell and harmonics of source voltage. Test results of proposed UPQC have produced acceptable results under balanced/unbalanced load conditions.