Implementation of a Deep Learning ANN-based Algorithm utilizing the IEEE 34 Bus Test System to Investigate the Effects of Distributed Generation on Fault Diagnosis in Distribution Networks

Abstract

Listen

Translate article

Electrical distribution systems are undergoing significant modifications since the application of new technologies. New possibilities for automated, dependable, and efficient electrical power grids have been made possible by the technological advancement. While new technologies might improve electrical network performance and offer creative solutions to future network difficulties, they can also have unintended consequences that need to be carefully studied and considered. A recent technological advancement that enhances power grid performance is Distributed Generation (DG). While DG unit integration has measurable benefits for electrical grids, its significant effects on power network protection systems create many questions and difficulties about the proper way to identify and isolate distribution network faults. The DLANN-based approach looks into the ways the integration of DGs affects fault identification and location. This method involves two steps: first, three-phase currents are constantly analyzed for detection, and Discrete Wavelet Transform (DWT) is utilized to extract the currents' features. The second step is classification employing Artificial Neural Networks (ANNs) to pinpoint the defective stages. Counting the shorted phases will reveal the sort of short circuit. The MATLAB programming environment is utilized in the development of the fault identification and classification technique. The fault type (one, two, or three phases), fault resistance, fault location bus, fault distance, and the DG type (upstream or downstream) are all considered. The methodology is used on a modified IEEE 34-bus test system, and four scenarios, one with combined DGs units, one with IBDGs, one with SBDGs, and one without DGs, are modeled. As per the simulation results, 100% fault detection and classification accuracy were obtained, whereas the average fault location accuracy attained without DGs, with IBDGs, SBDGs and combined DGs for selected nodes were 99.94%, 99.91%, 99.86%, and 99.88%, respectively