Condition monitoring of gas-turbine electric power units using the H-infinity Kalman Filter

Abstract

The article proposes a method for fault diagnosis and for detection of cyber-attacks in thermal power units that consist of a gas-turbine and a synchronous generator, after making use of the H-inflnity Kalman Filter. By performing approximate linearization in the dynamic model of the power system, through Taylor series expansion and through the computation of Jacobian matrices, the application of the H-inflnity Kalman Filter as a robust state estimator is possible. The H-inflnity Kalman Filter stands for a model of the fault-free functioning of the system and by subtracting from its outputs the outputs of the power unit the residuals’sequence is generated. Next, it is shown that the sum of the square of the residuals vectors being weighted by the inverse of the residuals’ covariance matrix, is a stochastic variable that follows the x2 distribution. Using the confidence intervals of the $\chi_2$ distribution one can define fault thresholds that indicate reliably the existence of a failure in the power system or the appearance of a cyberattack. Moreover, by applying the statistical test separately to the individual components of the power unit, fault isolation can be performed.