Bayesian Physics-informed Neural Networks for System Identification of Inverter-dominated Power Systems

TL;DR

The paper tackles the challenge of identifying power-system dynamics under increasing uncertainty from inverter-based resources. It extends Physics-informed Neural Networks with Bayesian inference (BPINN) to quantify both aleatoric and epistemic uncertainty and to estimate system parameters, using weakly-informative priors and SVGD-based variational inference. Across SMIB, 3-bus, 14-bus, and IEEE 118-bus grids with varying IBR penetration, BPINN substantially lowers identification errors compared with SINDy and outperforms PINN in uncertain, IBR-rich scenarios, while providing a principled uncertainty measure. Transfer learning from SMIB to larger grids significantly reduces training iterations and data requirements, with collocation points further aiding learning under data sparsity, highlighting practical scalability for real-time or near-real-time system identification in modern grids.

Abstract

While the uncertainty in generation and demand increases, accurately estimating the dynamic characteristics of power systems becomes crucial for employing the appropriate control actions to maintain their stability. In our previous work, we have shown that Bayesian Physics-informed Neural Networks (BPINNs) outperform conventional system identification methods in identifying the power system dynamic behavior under measurement noise. This paper takes the next natural step and addresses the more significant challenge, exploring how BPINN perform in estimating power system dynamics under increasing uncertainty from many Inverter-based Resources (IBRs) connected to the grid. These introduce a different type of uncertainty, compared to noisy measurements. The BPINN combines the advantages of Physics-informed Neural Networks (PINNs), such as inverse problem applicability, with Bayesian approaches for uncertainty quantification. We explore the BPINN performance on a wide range of systems, starting from a single machine infinite bus (SMIB) system and 3-bus system to extract important insights, to the 14-bus CIGRE distribution grid, and the large IEEE 118-bus system. We also investigate approaches that can accelerate the BPINN training, such as pretraining and transfer learning. Throughout this paper, we show that in presence of uncertainty, the BPINN achieves orders of magnitude lower errors than the widely popular method for system identification SINDy and significantly lower errors than PINN, while transfer learning helps reduce training time by up to 80 %.

Figures (11)

• Figure 1: Bayesian Phyiscs-informed Neural Network schematic with nonlinear activation $\eta$
• Figure 2: Exemplary normal-gamma PDFs for different parameters $\mu$, $\kappa$, $\alpha$, $\beta$
• Figure 3: 3-bus system with ibr
• Figure 4: CIGRE 14-bus MV system with ibr
• Figure 5: IEEE 118-bus with ibr