Ground fault insulation monitoring method for smart substation based on Mahalanobis distance and automatic code generation

Abstract

The existence of a ring network significantly increases the proportion of harmonic components in the DC system branch current, correspondingly diminishing the low-frequency components, resulting in a decrease in signal-to-noise ratio and negatively impacting the accuracy of ground fault insulation monitoring. Consequently, an intelligent substation grounding fault isolation monitoring method based on Mahalanobis distance and code automatic generation is proposed. Utilizing the characteristics of grounding faults, the complex wavelet method is employed for fault detection, effectively addressing the issue of inaccurate results caused by directly extracting low-frequency components when the branch harmonic content is high. Based on the detection results, the Mahalanobis distance algorithm is utilized for fault localization. Subsequently, monitoring software was designed using automatic code generation technology, combined with real-time monitoring of DC bus to ground insulation resistance and inspection of faulty branches, to achieve insulation monitoring of grounding faults. The experimental results demonstrate that the amplitude error of this method is as low as 0.03%, the phase error is only 0.15%, the relative error remains below 1.0%, the monitoring accuracy is as high as 92%, and the maximum monitoring response time is only 2 s, substantiating that this method exhibits excellent monitoring performance and can significantly reduce errors in the monitoring process.substantiating that this method exhibits excellent monitoring performance and can significantly reduce errors in the monitoring process.