Answers are grounded in this paper and cite their sources.
Simulation study demonstrates accurate state estimation using correlated Gaussian processes in power distribution grids, highlighting robust monitoring despite sparse sensor measurements.
Modern power distribution grids face significant monitoring challenges due to a lack of measurements. This results in limited insight into the state of the distribution system, which may cause issues regarding power system stability. In this study, we present a Distribution System State Estimation (DSSE) method that is data-driven and based on correlated Gaussian processes (Corr-GPs). The proposed approach is formulated without the use of pseudo-measurements and requires only 25 samples per node for training. Two variants have been developed: an individual model for estimating voltage magnitude and phase angle, and a coupled model that exploits cross-correlations. These variants have been evaluated on multiple time series and monitoring scenarios. On the IEEE 33-bus system with 11 monitoring units, the individual approach achieves an RMSE of 0.38 [10−2 p.u.] for voltage magnitude and 0.11 [10−3 rad] for phase angle. A slight degradation in performance is observed with a reduction to six measurement units. Compared to an artificial neural network, the proposed method achieves comparable accuracy while requiring orders of magnitude less training data. Scalability is demonstrated on the IEEE 69-bus system with 23 monitoring units, where estimation errors remain below 0.46 [10−2 p.u.] and 0.75 [10−3 rad]. These results demonstrate the potential of Corr-GPs as practical tools for state estimation in modern distribution grids.
No takes yet. Share an insight, caveat, or question.
Buechel et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: