A Fast State Estimation Technique using Deep Neural Network in a Distribution System

This study presents a fast state estimation technique using deep neural network (DNN) for voltage profile in a distribution system. While Monte Carlo Simulation (MCS) provides high accuracy in handling uncertainties of loads and variable renewable energy sources (VRE), its high computational cost remains a major challenge. To address this issue, the proposed method applies time-dependent and node-specific data preprocessing, enabling the DNN to maintain high inference accuracy under various network conditions. A case study using a modified IEEE 33-bus system model shows that the proposed method improved mean absolute error (MAE) by approximately 97% compared to conventional models. The proposed method not only enhances computational efficiency but also ensures robust accuracy, offering a practical solution for real-time grid monitoring and probabilistic power flow analysis.