Linearly constrained adaptive algorithms for line-frequency signal processing

Linearly constrained LMS adaptive filter algorithms are considered for digital processing of 50/60 Hz line-frequency signals. The constraints are set such that the primary sinusoidal waveform is guaranteed to pass the filter unaltered, and the adaptation is used to dynamically optimize the noise attenuation properties. In order to reduce the computational complexity of the constrained algorithm, selective coefficient updating is used, and the update formulas are derived accordingly. The approach is efficient in suppressing noise and harmonics in applications such as reactive power estimation and zero-crossing detection.