Improving Power System State Estimation through Physics Informed Deep Learning using Gated Recurrent Units

Over the past decade,renewable energy integration has transformed power systems, necessitating accurate state estimation amidst increased uncertainty.This paper aims to improve and enhancepower system state estimation by introducing a novel approach utilizing Deep Neural Networks (DNNs), including Gated Recurrent Unit (GRU). Inspired by autoencoders, the Physics InformedDeep Learning(PIDL) model integrates DNNs including Long Short Term Memory (LSTM), GRU, and Feed-Forward Neural Networks (FFNN), comparing them with both the traditional Weighted Least Squares (WLS) state estimatorand other versions of PIDL estimators.The estimated system states undergo validation against power flow (physical) equations to ensure compliance with fundamental laws. Therefore, the PIDL approach combines data-driven principles with physics guidance.Simulations conducted on the standard IEEE-14 and IEEE-30 bus systems in MATLAB demonstrate superior performance in terms of accuracy, robustness, and runtime speed. Furthermore, a comparative analysis between the proposed PIDL model and a counterpart from published literature underscores its efficacy and contributions to advancing cutting-edgePIDLpower system state estimation.