Multiresolution Signal Decomposition With Haar Wavelets for High‐Precision Power‐Frequency Estimation

ABSTRACT

Accurate and low‐latency estimation of power‐system frequency is critical for protection, control, and monitoring, particularly in modern grids with high renewable penetration and reduced inertia. Under faulted and nonstationary conditions, conventional methods—especially zero‐crossing–based techniques—exhibit significant inaccuracies due to decaying DC offsets, harmonics, and transient disturbances. This paper proposes a novel frequency estimation approach based on Haar wavelet multiresolution signal decomposition (MRSD), where frequency is inferred from local slope reversals (extrema) rather than zero crossings. A sinusoidal prefilter enhances noise robustness, while the multiresolution framework isolates the fundamental component from harmonics and disturbances. The proposed estimator provides both frequency and ROCOF estimation with low computational complexity and fast dynamic response. Its performance is validated using both simulation scenarios and real measurement data, including PMU‐based datasets and representative grid disturbance events. A quantitative comparison with conventional DFT‐ and PLL‐based estimators demonstrates improved accuracy, reduced oscillatory behavior, and faster transient tracking. Owing to its model‐free operation, robustness to distortions, and linear computational complexity, the proposed method is well suited for real‐time implementation in digital protection devices, PMUs, and power quality monitoring systems in low‐inertia power networks.