Development of an Intelligent Protection Coordination System for Overcurrent and Ground Fault Relays Using Neural Network Methods

Abstract

PT PAMA PERSADA, a prominent power plant located in Barunang, Kalimantan, operates with two primary feeders. However, persistent issues with protection coordination have been observed in one of the feeders, leading to disruptions in the plant's operational performance. The root cause of these issues was identified as improper coordination of overcurrent relays and ground fault relays. To address this, an evaluation of relay setting coordination was conducted using Electrical Transient Analysis Program (ETAP) 19.0.1 software. Specifically, the disturbances were concentrated in Feeder 2, which supplies power to the KM30 and Parilahung substations, where relay coordination failures resulted in simultaneous tripping events. Through ETAP simulations, the system's power parameters were analyzed, revealing active power of 4.164 kW, reactive power of 780.3 kVAR, apparent power of 4.237 kVA, and a power factor (cos phi) of 0.9. The protection coordination settings adhered to the IEEE 60255 standard, with grading times maintained between 0.2–0.6 seconds. To further enhance the reliability and optimization of the relay settings, simulations were extended using MATLAB software. This approach incorporated the Neural Network (NN) method for fine-tuning relay coordination, achieving a Mean Square Error (MSE) value of 0.33452, demonstrating significant improvements in fault management and protection system efficiency. These findings underscore the importance of integrating advanced computational tools and machine learning techniques for optimizing protection systems in modern power grids, contributing to improved operational reliability and adherence to international standards.