Evidential KNN-based condition monitoring and early warning method with applications in power plant

It is essential and challenging to monitor complex industrial processes and thus make an early warning for abnormal conditions, in particular when no fault samples can be observed under unknown uncertainties. To solve this problem, this paper proposes a so-called CMEW-EKNN method, i.e., condition monitoring and early warning method based on the evidential k-nearest neighbor (EKNN) rule in the framework of Evidence Theory. By employing the distance reject option in the EKNN rule, only normal operating data is needed to construct the early warning model. An adaptive discounting factor is adopted to make the early warning boundary adaptive to local distribution characteristics of the training samples, so as to improve both effectiveness and robustness of CMEW-EKNN. Comparisons on two practical applications in power plant demonstrate that the proposed CMEW-EKNN, which adopts the adaptive discounting factor, yields superior fault early warning performance than the PCA-based and FD-kNN fault detection approaches.