Robust State Estimator Based on Conditional Variational Autoencoder

Junjian Qi,Lei Wang

DOI: https://doi.org/10.1109/PESGM51994.2024.10688686

2024-07-21

Abstract:In this paper, we propose a novel, model-free robust state estimator based on the conditional variational autoencoder (CVAE). Unlike traditional model-based methods such as weighted least squares, our approach eliminates the need for prior knowledge of the power system model. Once adequately trained, our estimator can accurately estimate the system states based on measurements. As CVAE provides a probability distribution over the output data, the inherent uncertainty modeling makes the proposed estimator robust to bad data. The effectiveness and robustness of the proposed state estimator are demonstrated on the IEEE 14-bus system.

Computer Science,Engineering