State Estimation Method for Distribution Network Based on Incomplete Measurement Data

With the large-scale integration of distributed energy resources, the operational characteristics of the traditional distribution networks have undergone significant changes, leading to such problems as dispersed loads, poor real-time observability, and incomplete data, which severely impact the state monitoring and operational optimization of the distribution networks. To address above problems, we propose a distribution network state estimation method based on Bayesian-optimized convolutional neural networks (CNN) and long short-term memory (LSTM) networks with incomplete real-time measurement data. The method is divided into two phases: offline learning and online state estimation. In the offline learning phase, generative adversarial networks are used to generate the required samples for training the CNN-LSTM model, and the Bayesian optimization algorithm is employed to adjust the hyperparameters, thereby enhancing the accuracy of the algorithm. In the online state estimation phase, the state estimation is performed online with incomplete real-time data of the distribution network and the trained CNN-LSTM model. Finally, simulation analysis is conducted on the IEEE 33 and IEEE 123 networks, which confirms the effectiveness and accuracy of the proposed state estimation method.