.png){kind=icon}
.png){kind=icon}
Return
Approximate Bayesian Computation Guided Sequential Neural Posterior Estimation for Power System Parameter Calibration
.png){kind=icon}
PRE
AI
DOI：10.1109/TPWRS.2025.3647318.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Generator parameter calibration is essential for power system analysis and control. With intractable likelihood function due to complex dependencies between parameters and the simulation outputs, simulation-based methods such as approximate Bayesian computation (ABC) and sequential neural posterior estimation (SNPE) are popular approaches for parameter estimation. In this paper, we propose a novel ABC-guided SNPE model for generator parameter calibration by integrating ABC and SNPE to leverage the strengths of both techniques. The initial screening with ABC quickly rules out the implausible regions of the parameter space and a normalizing flow based generative model then transforms the coarse posterior approximation from ABC into a smooth, continuous distribution that provides better quality training data for SNPE. Our approach is validated on a synchronous generator model and the results demonstrate the advantages of our approach over ABC or SNPE alone in both estimation accuracy and computational efficiency.
Keywords:
Approximate Bayesian computation (ABC)
efficiency
normalizing flow
parameter estimation
sequential neural posterior estimation (SNPE)
simulation-based method
synchronous generator dynamic model
Journal
IF:
7.2
Papers:
1.1W
Citations:
5.0W