Research on power distribution network state estimation algorithm based on deep integration of LSTM and extended Kalman filter

In the context of managing and operating distribution networks, state estimation serves as a vital tool for gaining comprehensive insights into the state of network and constructing real-time digital twins. The precision of this estimation directly influences the optimization of grid scheduling and the formulation of control decisions. However, the measurement data used in state estimation often encounters quality challenges, undermining the accuracy and reliability of conventional methods. Addressing this, the paper introduces an advanced algorithm that merges Long Short-Term Memory (LSTM) networks with the Extended Kalman Filter (EKF). By employing LSTM, the model extracts temporal and network-dependent features from measurement data, thereby compensating for data quality deficiencies and robust model. Meanwhile, the EKF effectively tackles the nonlinear attributes of the distribution network, minimizing the detrimental effects of noise and system disruptions. Through validation using the IEEE 33-bus system, the LSTM-EKF algorithm has shown remarkable improvements in accuracy and robustness. This research not only presents a potent algorithmic framework to address data quality concerns in state estimation but also supports the digital grid in accurately capturing and reflecting the real-time state of the physical network.