Neural Network Based Control Algorithm for Solar PV Interfaced System

With rising weak utility grid conditions, the power quality degradation has proved to be a major cause of concern due to the escalation in integrating renewable energy sources with the distribution grid. Thus, for power quality (PQ) enhancement, it is paramount to implement a robust control technique for the grid interfaced solar energy conversion system. A neural network (NN) based gradient descent with momentum control technique is presented here for performing functions of distribution static compensator (DSTATCOM) such as correcting power factor, load balancing and alleviating harmonics for optimal operation along with transferring active power to the load and the grid. The weights are adjusted adaptively through the usage of the control algorithm in an independent manner with reduction in computational time and hence during abnormal grid conditions the model complexity is reduced. Moreover, the estimation process utilizes neural network which enhances accuracy and reduces computational burden. For validating the behavior of proposed system under non-ideal conditions of the grid, a prototype developed in the laboratory is tested and the results corroborate reliable operation.