Return
A Data-Driven Measurement Uncertainty Prediction for the Smart Charger Meters Based on a Posterior Spatio-Temporal Network
DOI:10.1109/TII.2025.3621484.png)
Abstract
En 中文
The measurement performance of electric meters is significant for the safe and reliable operation of electric chargers. In this article, an online prediction method is proposed based on a posterior spatio-temporal network to evaluate the measurement uncertainty of the electric meters of chargers. Compared with the relative error, which only reflects the accuracy of the meter measurement, the measurement uncertainty can reflect both accuracy and precision. First, a posterior LeNet block is designed by integrating the fully connected layer with Bayesian theorem to capture the mapping between electricity parameters and measurement uncertainty. In addition, to learn the dynamic changes in timewise uncertainty of meter data, the obtained spatial characteristics are fed into a posterior gated recurrent unit (GRU) block where candidate hidden states follow a posterior distribution instead of fixed weights. By this way, the change trend of measurement uncertainty can be captured through the information transmission between units. To compute the gradient information of the distribution during the backpropagation process, the reparameterization technique is employed. Finally, the proposed method is verified in an actual charging station. Compared with the state-of-the-art methods, experimental results demonstrate the superiority of the proposed method in the prediction of measurement uncertainty.
Keywords:
Bayesian theorem
measurement uncertainty prediction
posterior spatio-temporal network (POSTN)
smart charger meter
Journal
IF:
9.9
Papers:
8.8K
Citations:
6.0W
Organization
Cited Papers
Layer adaptive node selection in Bayesian neural networks: Statistical guarantees and implementation details
NEURAL NETWORKS
IF6.3
A measurement error prediction framework for smart meters under extreme natural environment stresses

