A Robust Fault Detection Strategy of Generator Units Based on EWMA Residual Filter

Model-based residual directly affects the performance targets of fault detection system. An effective means of evaluating the residuals is the appropriate filter to restrain uncertain disturbances. This paper proposes a residual-based exponentially weighted moving robust fault detection strategy, and gives a EWMA (exponentially weighted moving average) robust early alarm process. Applied in water leakage fault detection of generator units, fault detection system performs well in both fast detection ability to abrupt fault and the sensitivity to the slow fault, and has good robustness.