A DGCRN-based harmonic source localization method for distribution systems with network uncertainty

• The DGCRN-based method locates harmonic sources in distribution systems with network uncertainty. • Dynamic and static topologies are integrated for robust spatiotemporal feature extraction. • The HHSL architecture enables real-time source localization with limited harmonic RMS data. • The proposed method achieves high localization accuracy with low computational requirements. Accurate localization of harmonic sources is crucial for determining harmonic responsibility and mitigating harmonic distortions. However, uncertainties introduced by network switching operations, inaccurate line parameters, and power flow variability often result in low localization accuracy for harmonic sources, as existing methods rely on precise topology and synchronous phasor information that are often unavailable in practice. To address these challenges, a harmonic source localization method based on a dynamic graph convolutional recurrent network (DGCRN) is proposed for locating harmonic sources in distribution systems with network uncertainty. Firstly, a network uncertainty representation method is proposed for distribution systems, which incorporates a static topology reflecting electrical distances and a dynamic topology capturing the time-varying harmonic couplings between nodes. Secondly, a hierarchical harmonic source localization (HHSL) architecture is proposed, which integrates multi-layer data. Accurate harmonic source localization at the data center is achieved by employing a reconstructed DGCRN that leverages low-dimensional key features. Compared with existing methods, the proposed method exhibits reduced reliance on predefined topology and requires only harmonic RMS values as feature data. The modified IEEE-33 test system and the actual distribution system in South China are both tested to validate the effectiveness of the proposed method.