Voltage Signal-Based Intelligent Fault DetectionConsidering Reactive Power Injection from Inverter-Based DGs

Voltage signals can be good candidates for designing protection schemes owing to the characteristics of sag in faulty phase(s) and swell in healthy phase(s). However, many grid codes have provision for reactive power injection from inverter-based distributed generation units integrated in power system networks in order to support the network voltage levels during the fault periods. As such, the voltage levels get affected due to this provision, which in turn may impact the fault detection capability of the protection schemes designed based on the voltage signals. This paper presents an intelligent fault detection scheme based on features extracted from the local voltage signals using empirical wavelet transform. The extracted features are used to develop a fault detection scheme based on machine learning approach. The study is performed on the IEEE 13-bus unbalanced system with inverter-based distributed generation units modeled in MATLAB. Various scenarios, especially those related to the reactive power injection along with varying penetration of the distributed generation units, switching events and noise in signals, are considered. The proposed work shows encouraging performance under such diverse cases leading to a scheme which is accurate, reliable and easy to implement.