Impact of Pseudo-Measurements Generation on Distribution System State Estimation

The evolving structural changes in power networks have a significant impact on the management of monitoring and control applications. Among them, Distribution System State Estimation (DSSE) faces inherent limitations due to uncertainties arising from these transformations, which often lead to a degradation in the quality of measurements and pseudo-measurements used in state estimation routines. To mitigate these challenges, Machine Learning techniques are increasingly recognized as effective solutions to improve the performance of monitoring applications. In this context, this paper aims to assess how the prediction of active and reactive powers obtained through a Multi-Layer Perceptron (MLP) neural network and compared with simple benchmark models, affects DSSE performance. Firstly, starting from real data collected from the Forschungszentrum Jülich campus, the MLP model has been characterized and, finally, DSSE has been evaluated by means of several numerical simulations. The preliminary exploratory results have suggested that the proposed model shows promising potential in improving the accuracy of DSSE. These initial results suggest that it may be worth investigating more complex neural models in the future, with the aim of further enhancing DSSE performance and providing system operators with increasingly reliable monitoring and control tools.