Comparative Analysis of the Effect of KPCA in PSO-GRU Combination Model

Considering the issues of strong coupling, complex fault mode and nonlinear relationship between variables in marine condensate feed water system, the kernel principal component analysis (KPCA) is suggested as it offers significant benefits in addressing nonlinear problems. Combined with gated recurrent unit (GRU), which can improve the advantages of gradient explosion phenomenon in dealing with complex timing problems. The KPCA-PSO-GRU combined model is constructed for fault classification prediction. Samples were constructed from typical faults in the operation of the condensate feed water system, and the main features of monitoring parameters were extracted by kernel principal component analysis (KPCA) to reduce the dimensions of the original data, which were then input into the PSO-GRU model, and the original data without KPCA were directly input into the GRU model for comparative analysis. The results show that KPCA has greater advantages in extracting features from nonlinear data, and the KPCA-PSO-GRU combined model has higher accuracy in predicting fault classification, which can shorten the time for fault prediction and diagnosis of condensate feed water system, and has certain guidance.