Combination of wavelet transforms and artificial intelligence in HV shunt capacitor banks: Fault and inrush currents classification

The installation of HV substation shunt capacitor bank in a substation and transmission system affects both the capacitor inrush currents and fault currents. This is led to mal-operation of the overcurrent protection relay (50/51) and unbalance current protection relay (60C, 51NC) to classify whether it is the inrush current or fault current. The protective relay may be tripped due to transient inrush currents during a capacitor energizing. This paper presents a new detection and classification technique by using a combination of the discrete wavelet transform (DWT) and probabilistic neural network (PNN). The DWT on the measured three-phase current signals was employing the effective time-frequency analysis for detection. The detail wavelet coefficients of level-1 to level-30 of currents signals are extracted by DWT using 'db4' mother wavelet. The suggested PNN technique uses to classify those transient. The two classification techniques such as 1) only DWT and 2) combined DWT and fuzzy inference system (FIS) are compared with the proposed method. The performance of the proposed technique is found to be accurate for the detection and classification of power system transient in a capacitor bank which it has high accuracy by 100% for detection and more than 98% for classification.