Smart Inverters (SIs) have the capability to regulate the distribution grid voltage at the point of common coupling (PCC) through the injection or absorption of reactive power. The operating conditions of a distribution grid are time-varying due to the fluctuation of distributed energy resource (DER) output and load demand. Therefore, static or one-time setting of SIs' droops may not work for maintaining the nodal voltages on a distribution feeder over a duration of time. In this context, this work puts forward a Neural Network (NN)-based approach to dynamically adjust droop settings of SIs in response to the variation of DER power and load demand. The performance of NN is demonstrated using the large-scale IEEE 8500-node test feeder. The predicted SI droop settings align close with the ground truth values, which highlights the usefulness of the proposed adaptive droop setting for effective voltage regulation on Active Distribution Networks (ADNs) as the operating condition varies.
