A multi-layered framework for stability analysis in power grids with incomplete harmonic data

In modern distribution network incomplete and noisy harmonic measurement from smart meters and grid sensors can degrade the power quality monitoring and mislead stability assessment. This research presents a multi-layer framework that reconstructs harmonic signals under missing data and evaluates the dynamic stability impact of reconstruction-induced perturbations. First, adaptive wavelet-based denoising and windowed segmentation are applied to suppress noise while preserving harmonic content. Next, a lightweight reconstruction strategy inspired by autoencoders is implemented using ridge regression to impute missing samples with low computational cost. Signal structure preservation is examined using dual-scale visualization via PCA and t-SNE. System-level effects are then assessed through eigenvalue trajectory mapping and a proposed Bifurcation Stability Margin (BSM) metric that quantifies sensitivity to imputation errors, supported by reduced-order modelling and harmonic sensitivity analysis for scalability. Results show high reconstruction accuracy (MAE = 0.0022), while stability analysis indicates that even small harmonic inconsistencies can cause noticeable eigenvalue drift (λ_max = 1.2847) and increase bifurcation risk (BSM = 1.3371). These findings demonstrate that accurate signal recovery must be coupled with stability-aware validation, and the proposed framework provides a practical pathway for real-time, resilient power-quality monitoring and early-warning support in smart grid.