State Estimation for Distribution Network Based on Graph Neural Network Method

With the continuous increase in urban electricity consumption, the topology scale of distribution networks is becoming larger and the distribution network environment is becoming more diverse, resulting in some of the collected data containing noise. Existing distribution network status assessment methods face many problems in this regard. This paper proposes a distribution network state estimation method based on graph neural networks, which fully utilizes the correlation characteristics between distribution network data extracted by graph neural networks, excavates the causal relationship between node active power, reactive power and distribution network state variables, and achieves accurate evaluation of distribution network state under noise conditions. Finally, it is demonstrated that the proposed method has good fault tolerance in the presence of noise by adding noise to the IEEE33 and IEEE69 node systems.