Neupane, A.Paudyal, S.Ceylan, O.2026-01-152026-01-15202597814673272759781538677032979835038183297814799130399781665405072978146738040997815090416889781728119816978172815508197814799641541944-9925https://doi.org/10.1109/PESGM52009.2025.11225698https://hdl.handle.net/20.500.12469/7696Smart Inverters (SIs), which are power-electronics based devices, have capability to effectively regulate the voltage on distribution feeders with better time granularity due to their faster response compared to legacy devices such as on load tap changers (OLTCs) and capacitor banks. In this study, we propose an approach to predict the droop settings of SIs that dynamically adjust based on network conditions as observed through the voltage measurements. This work adopts Long Short Term Memory (LSTM) based Neural Network (NN) approach for predicting dynamic droops for SIs as the network condition changes. We test the effectiveness of the dynamic droops on a large (IEEE 8500-node) distribution network. The case studies demonstrate that the voltage performance on distribution feeders can be improved with dynamic droop settings. © 2025 IEEE.eninfo:eu-repo/semantics/closedAccessDistribution GridDroop SettingLSTMSmart GridsSmart InvertersConference Object10.1109/PESGM52009.2025.112256982-s2.0-105025200907