Abstract: In the process of sharing power data through online channels, there exists a potential risk of data leakage, necessitating effective data traceability of leaked information. This paper proposes a novel method for tracing semi-structured power data based on robust zerowatermarking, which enables traceability even when the leaked data undergoes local modifications. First, relevant key-value pairs are extracted from the semi-structured power data to create a feature sequence as the carrier for the zero watermarking. Then, the feature sequence and the watermark body are divided into equal-sized blocks. By integrating erasure codes and redundant error correction codes theory, an intermediate sequence is obtained through a transfer matrix, and then encoded using error correction codes. Finally, the error corrected watermarking information is embedded into the feature sequence, generating a robust zero watermarking for the power data. During the tracing process, the robust zero watermarking of the tracked data is extracted and analyzed to effectively identify and locate data anomalies. Experimental and simulation results demonstrate that the proposed method achieves a watermark extraction success rate of over 98%, while ensuring data security. The presented approach holds significant application value in the monitoring and anomaly tracing of power systems' data.