A Method for Power Flow Estimation in Power Systems with Incomplete Data

Machine learning-based power flow estimation is a critical technique for modeling the operational state of power systems. However, traditional power flow estimation methods face limitations under conditions of low-quality measurements. The challenge of accurately performing power flow estimation with missing data has become a key area of research. This paper proposes a convolutional autoencoder model to enhance power flow estimation in power systems. The convolutional layers effectively capture local dependencies in the power flow data, while the autoencoder extracts the intrinsic features of the flow, which are then used to repair missing data. Validation on a large dataset demonstrates the model's ability to restore power flow data across varying missing data rates, showing excellent accuracy and stability.