An intelligent ANFIS-based fault detection, classification, and location model for VSC-HVDC systems based on hybrid PCA-DWT signal processing technique

Abstract

Designing an accurate model for fault detection, classification, and location is vital from the protection viewpoint. The adaptive neuro-fuzzy inference system (ANFIS) is a commonly used learning-based fault location model that performs independently from the propagating wave characteristics, whose performance can be improved by optimizing the membership functions associated with the inputs. Accordingly, two metaheuristic algorithms with the quick-search capability of the population space, i.e., Harris Hawks optimization (HHO) and cuckoo search (CS) optimization algorithms, are used in this paper to optimize the ANFIS, and their performances are compared with the traditional ANFIS training algorithms. Moreover, principal component analysis (PCA) is employed to detect the fault, and the discrete wavelet transform (DWT) strategy is exploited to acquire the ANFIS training and testing dataset according to the statistics T² obtained by PCA. Three statistical indices, i.e., mean value, standard deviation, and norm entropy, are computed corresponding to the extracted wavelet coefficients from the current signal and applied to train and test the ANFIS. Optimized ANFIS conducts fault classification and location tasks, and the accuracy of the proposed model is compared with the commonly used traveling wave (TW) –based models and recently proposed fault location methods in the literature. Three fault types on the DC-link with three fault resistances are examined to confirm the fault classification superiority of the proposed model and its fault type/impedance-independent estimation capability. An accuracy rate of 99.995% is obtained for the fault locating task, while fault detecting and classifying are accomplished with an accuracy of 100%. Simulations and numerical studies are performed in MATLAB software.