Neural Network-Based Fault Distance Estimation for Multi-Terminal DC Microgrids

Fault distance estimation in DC microgrids is critical due to the growing adoption of DC-based distribution systems. Current methods face limitations like sensitivity to system parameters and high-resistance fault detection, necessitating improved accuracy. This study proposes a neural network approach to locate fault distances in multiterminal DC microgrids. Three different structures based on back propagation algorithms are developed and trained to accrately estimate fault distances with high precision. These structures can handle various fault scenarios, including different fault resistances and the presence of noise. Two structures can predict fault distances from one side locally, achieving low error rates of 0.3 % for the source side and 0.6 % for the load side. The third structure incorporates input variables from both sides, resulting in even more accurate predictions with an error rate of less than 0.15 % for both terminals. A comparative analysis was performed to evaluate the proposed fault distance estimation structures regarding error percentage, cost, fault resistance, and reliance on communication systems. The results demonstrated the superiority of the proposed structures in all aspects, emphasizing their effectiveness in improving the performance of the protection system.