Enhanced RF for Fault Detection and Diagnosis of Uncertain PV systems

This paper proposes a new fault detection and diagnosis (FDD) approach based on Random Forest (RF) for uncertain Grid-Connected PV (GCPV) system. Firstly, to deal with uncertainties, interval RF based on upper and lower bounds of interval matrix RFUL is proposed. To more improve the diagnosis accuracy, interval kernel PCA (IKPCA)-based RF classifier is developed. The enhanced scheme is so-called IKPCA-based RF and it is based on three major phases: feature extraction (FE), feature selection (FS) and fault classification (FC). The basic idea behind FE and FS phases is to use IKPCA model in order to select the most relevant and informative features from raw data. IKPCA aims to fit two KPCA models on the lower and upper bounds of the variables interval values. Then, the sensitive and significant interval-valued characteristics are transmitted to the RF model for classification purposes. The presented results demonstrate the effectiveness of the proposed interval-valued methods when applied to uncertain GCPV systems.