Fault detection based on two-stage recursive least squares parameter estimation

This paper proposes a two-stage recursive least squares algorithm for output error models. systems are considered multiple-input single-output, which is represented by autoregressive models. The first step of the proposed algorithm is to identify algorithm model. Second one is to implement these identification methods at thermal power plant, whose nominal power output is 320 MW. The water level measurements are estimated by two-stage recursive least squares parameter estimation. To validate the model, autocorrelation function of measurement residuals and cross-correlation function between input signals and residuals are analyzed. next, the determination of residual and the mean value of the residual divided to the standard deviation of the residual of two-stage are used for fault detection. Comparison between the conventional model recursive least squares algorithm and two-stage recursive least squares algorithm is finally presented.