Improved Recursive Newton Type Algorithm based power system frequency estimation

Frequency measurement approach based on linear model for increasing low SNR sinusoidal signal frequency measurement precision

“…The fitting equations of all method are expressed as (39)-(41). According to these equations, we can know that, the relative error of the proposed method is inversely proportional to the 1.467 power of the measurement time length, which is the same as the theory result expressed in (33). But for the MSCM and MPM, the relative error is only inversely proportional to the measurement time length δ MCSM = 3 × 10 −7 T 0.996 (39)…”

“…13a and b, which indicate that (i) the RMSD of the proposed method is smaller than MPM in the frequency range from 1 to 80 kHz; (ii) when frequency increases, the RMSD curves of both method increase in inverse proportion curve, and the RMSD of the proposed method rise rapidly as frequency increasing when the frequency is larger than 40 kHz. The phenomenon can be explained from (33): the length of a [i] (N in (32)) is set to 1000 for the last experiment, when the measured frequency increases, the time length of once measurement (MT in (31)) decreases, and according to (33), the relative error decreases correspondingly. In order to reducing the relative error for this situation, the MT must be increased, the measured frequency must be decreased correspondingly, so the divider of capture channel one on STM32 chip is adopted to divide the measured frequency when the frequency is larger than 10 kHz.…”

“…The relative error of (32) can be expressed as (33), which indicate that the measurement time length is longer, the relative error is smaller. The measured frequency is larger, the relative error is larger.…”

“…Thus the application of these methods in a portable device will be limited. There are some studies on power system frequency estimation, these include leastsquare (LS) techniques and their variants [25,26]; filtering method including notch [27], Kalman [28] and adaptive [29,30]; the DFTbased frequency estimation algorithms [31]; iterative approaches [32], Newton type algorithm [33] and so on. The frequency estimation precision is good enough because the estimated frequency is almost known.…”