2nd Order Linear Prediction Applied to Digital Frequency Estimation for Different Disturbances

To maintain the quality of energy supply and the correct operation of equipment, the electrical frequency is a parameter of high importance in Electric Power Systems. This work presents a digital frequency estimation technique based on network voltage signals analysis that are decomposed into their modal domain using the Clarke Transformation. Second order linear prediction is used in order to predict the future values of these components. In addition, a logic to reduce the prediction error is applied to improve the estimation response. The network frequency is given by the angle resulting from the multiplication between the complex signal and that given by
$\alpha$
and
$\beta$
components prediction. The technique was tested for cases with ramped, exponentially and damped sinusoidal frequency variations, as well as in the presence of White Gaussian Noise, and evaluated in terms of convergence time, minimum and maximum errors before and after convergence, showing that this innovative method has great precision and robustness in different simulation situations.