Eliminating bias in frequency estimation based on Prony's method

The fundamental frequency is an important parameter required for efficient power system control and protection functions. Prony's method gives fast and accurate frequency estimation when input signal has very high signal to noise ratio. For practical measurements when signal to noise ratio is low, frequency estimation based on Prony's method is very sensitive to noise and results are biased. This paper proposes a modified Prony's method for frequency estimation that eliminates bias and produces very accurate estimates. The proposed algorithm adds some computational burden because a few additional iterations are required to correct initial biased result produced by the classical Prony's method. This paper concludes with simulation and practical examples that are demonstrating application of the unbiased modified Prony's method for frequency estimation.