Abstract：With the accelerated construction of new power systems，the scale and complexity of transmission

systems are constantly increasing. Therefore，it is urgent to study transmission line fault diagnosis algorithms that

utilize multi-source data as driving sources and meet requirements for accuracy and low time consumption. A multisource information fusion transmission line fault diagnosis method based on the improved NRBO-XGBoost

algorithm was proposed. Firstly，by analyzing the measured electrical quantities and action switch quantities on

both sides of the line protection，the correlation features of time/frequency domain differential current and

differential voltage，transient polarity，and action signals under internal and external fault scenarios were decoupled.Then，the decoupled multi-source fault feature vectors were input into the XGBoost serial learning algorithm，and the NRBO algorithm was introduced to globally optimize the training parameters of XGBoost. Finally，based on the identification output of the improved NRBO-XGBoost algorithm，a complete transmission line fault diagnosis model for internal and external faults was obtained. An IEEE-30 standard node transmission system model was constructed using PSCAD/EMTDC. Through testing in four typical scenarios，the results demonstrated that the proposed multi-source information fusion algorithm achieves a line fault diagnosis accuracy of 99%，meeting the required threshold. Additionally，it exhibits certain advantages in terms of diagnosis speed compared to traditional intelligent algorithms.

      Key words：transmission line；fault diagnosis；XGBoost algorithm；NRBO algorithm；multi-source information

fusion