Enhanced fault detection and classification in monopolar HVDC transmission lines using MODWT and AI technique

Abstract

It is necessary to protect high-voltage direct current transmission lines (HVDC-TLs), and in order to guarantee continuous operation, faults must be detected quickly. In this paper, a suggested method for HVDC-TL system protection is presented. Detecting the short circuit and open circuit conditions is the first of the two sections of the study. The second step is to classify the short circuit as an external (AC network) or internal (HVDC line) issue. This paper uses the five-level maximal overlap discrete wavelet transform (MODWT) technique to analyze only the current signals in the steady-state and faulted states from the beginning of the transmission line. It then finds a polynomial curve fit and determines the R-squared and dynamic time warping (DTW) values. An adaptive neural fuzzy inference system (ANFIS) is trained to classify the fault using the R-squared and DTW values as input. This paper studies a monopolar HVDC-TL system built in PSCAD/EMTP and implements the proposed algorithm using MATLAB software. The suggested approach shows a strong ability to identify, classify, and locate faults in HVDC transmission systems, attaining up to 98.71% accuracy even in the presence of significant fault resistance and measurement noise. Furthermore, the method demonstrates excellent flexibility in a variety of HVDC topologies, such as bipolar and HVDC cable, and medium voltage DC transmission line, establishing its viability for real-world implementation. Although the approach is very reliable, it is sensitive to very high noise levels. A lightning fault simulation was conducted to evaluate the system’s performance under various surge magnitudes; 10 kA, 50 kA, and 100 kA. The results confirm the robustness and accuracy of the proposed method across all tested lighting conditions. • Fast detection of HVDC transmission faults. • Identifies short and open circuit conditions. • Classifies internal versus external HVDC faults. • MODWT and ANFIS based fault analysis. • Achieves 98.71% accuracy under noise conditions.