Detection and Imputation based Two-Stage Denoising Diffusion Power System Measurement Recovery under Cyber-Physical Uncertainties

TL;DR

This work tackles power-system measurement recovery under cyber-physical uncertainties such as FDIA and data losses, modeling both measurement and system dynamics with nonlinear, renewables-driven fluctuations. It introduces an enhanced two-stage denoising diffusion model (TSDM) comprising a Stage 1 classifier-guided conditional diffusion for anomaly detection and a Stage 2 diffusion-based imputation for missing data, augmented by subsequence sampling with an optimal variance to accelerate generation. The approach achieves high accuracy and robustness on SCADA and WAMS data across IEEE test systems, outperforming LSTM, VAE, GAN, and ADMM-based methods while reducing computational demands. Case studies demonstrate strong performance under diverse cyber-attacks and nonlinear dynamics, with practical implications for real-time cyber-resilient state estimation and measurement recovery in modern power grids.

Abstract

Power system cyber-physical uncertainties, including measurement ambiguities stemming from cyber attacks and data losses, along with system uncertainties introduced by massive renewables and complex dynamics, reduce the likelihood of enhancing the quality of measurements. Fortunately, denoising diffusion models exhibit powerful learning and generation abilities for the complex underlying physics of the real world. To this end, this paper proposes an improved detection and imputation based two-stage denoising diffusion model (TSDM) to identify and reconstruct the measurements with various cyber-physical uncertainties. The first stage of the model comprises a classifier-guided conditional anomaly detection component, while the second stage involves diffusion-based measurement imputation component. Moreover, the proposed TSDM adopts optimal variance to accelerate the diffusion generation process with subsequence sampling. Extensive numerical case studies demonstrate that the proposed TSDM can accurately recover power system measurements despite renewables-induced strong randomness and highly nonlinear dynamics. Additionally, the proposed TSDM has stronger robustness compared to existing reconstruction networks and exhibits lower computational complexity than general denoising diffusion models.

Figures (15)

• Figure 1: Random and non-random data losses distribution of received measurements in power systems.
• Figure 2: Measurement recovery challenges under irregular $\Delta \bm{P}_G$ and nonlinear dynamics by $\Delta \bm{u}$.
• Figure 3: Diffusion measurement prior model and its fast sampling method.
• Figure 4: Improved denoising diffusion implicit model with subsequence acceleration and parameterized reverse process $p_{\theta}$ under classifier-guidance.
• Figure 5: The flowchart of improved efficient TSDM with Stage 1: conditional anomaly detection and Stage 2: diffusion based imputation.