Data-Driven State Estimation Method for Distribution Networks Based on Power Flow Constraints

Addressing uneven accuracy across multiple measurement sources and the insufficient adaptability of traditional data-driven models to topological changes, a data-driven state estimation method for distribution networks based on power-flow constraints is proposed. Firstly, historical prior information is used to supplement missing data, thereby constructing a complete measurement dataset and establishing a reliability index to assess the credibility of measurements from different sources. Secondly, according to the physical topological map structure of the distribution network, an improved graph neural network algorithm is proposed to realize state estimation. The bus power balance relationship is embedded as a soft constraint in the loss function to enhance the consistency between the estimation results and physical laws. Then, an adaptive method of topology change based on transfer learning is proposed to improve the adaptability of the model to the dynamic changes of distribution network topology. Finally, the proposed method is evaluated using various metrics on the IEEE 33 and IEEE 118 systems. The analysis results show that the method can maintain high estimation accuracy under multi-source measurements and exhibits good robustness to noise disturbances.